Toward a Design Theory of Problem Solving

[] David H. J o n a s s e n

Problem solving is generally regarded as the [] Gagn~ believed that "the central p o i n t of e d u -
most important cognitive activity in everyday cation is to teach p e o p l e to think, to use their
and professional contexts. Most people are rational powers, to become better p r o b l e m solv-
required to and rewardedfor solving problems. ers" (1980, p. 85). Like Gagn~, most psycholo-
However, learning to solve problems is too gists and educators regard p r o b l e m solving as
seldom required informal educational settings, the most important learning outcome for life.
in part, because our understanding of its Why? Virtually everyone, in their e v e r y d a y a n d
processes is limited. Instructional-design professional lives, regularly solves problems.
research and theory has devoted too little Few, if any, p e o p l e are r e w a r d e d in their profes-
attention to the study of problem-solving sional lives for m e m o r i z i n g information a n d
processes. In this article, I describe d~erences completing examinations, yet examinations are
among problems in terms of their the p r i m a r y arbiter of success in society. Unfor-
structuredness, domain spec~city ttmately, students are rarely, if ever, required to
(abstractness), and complexity. Then, I briefly solve meaningful p r o b l e m s as p a r t of their cur-
describe a variety of individual d~erences ricula. The few problems that students d o
(factors internal to the problem solver) that encounter are n o r m a l l y well-structured (story)
affect problem solving. Finally, I articulate a problems, w h i c h are inconsistent with the
typology of problems, each type of which nature of the problems they will need to learn to
engages d~-erent cognitive, affective, and solve in their e v e r y d a y lives ( " H o w can I get so-
conative processes and therefore necessitates and-so to p a y attention to me?"), professional
d~j:erent instructional support. The purpose of lives ("What k i n d of m a r k e t i n g a p p r o a c h is
this paper is to propose a metatheory of appropriate for this n e w p r o d u c t line?"), or even
problem solving in order to initiate dialogue their school lives ("Should I s p e n d the next two
and research rather than offering a definitive hours studying for m y m a t h exam or go outside
answer regarding its processes. a n d p l a y ball w i t h m y friends?"). Therefore,
graduates are rarely, if ever, adequately pre-
p a r e d to function in e v e r y d a y a n d professional
contexts following education a n d training. The
discrepancy between w h a t learners need (com-
plex, ill-structured problem-solving experience)
I This paper represents an effort to introduce issues and and w h a t formal education (schools a n d corpo-
concerns related to problem solving to the instructional
design community. I do not presume that the community is rate training) p r o v i d e s represents a complex a n d
ignorant of problem solving or its literature, only that too ill-structured p r o b l e m that instructional design
little effort has been expended by the field in articulating m a y be able to ameliorate.
design models for problem solving. There are many reasons
for that state of affairs. W h y are w e so inept at engaging learners in
The curse of any introductory paper is the lack of depth in p r o b l e m solving? A major reason, I argue, is that
the treatment of these issues. To explicateeach of the issues
raised in this paper would require a book (which is w e do not u n d e r s t a n d the b r e a d t h of p r o b l e m -
forthcoming),which makes it tmpublishablein a journal My solving activities well e n o u g h to engage a n d
purpose here is to introduce these issues in order to stimulate s u p p o r t learners in them. Problem solving has
discussion, research, and development of problem-solving
instruction that will help us to articulate better design never been sufficiently a c k n o w l e d g e d or articu-
models. lated in the instructional design literature. W i t h

ETR&D,Vol.48, No. 4, 2000,pp. 63-85 ISSN1042-1629 63

 ETR&D, VcK 48, No. 4

few exceptions, it is not even mentioned in most rary conceptions of student-centered learning
textbooks on instructional design. To their environments, such as open-ended learning
credit, Smith and Ragan (1999) included a chap- environments (Hannafin, Hall, Land, & Hill,
ter on problem solving, however they prescribe 1994; Land & Hannafin, 1996), goal-based sce-
only general problem-solving strategies as solu- narios (Schank, Fano, Bell, & Jona, 1993/1994),
tions. GagnG Briggs, and Wager (1992) acknowl- and even problem-based learning (Barrows,
edged that problem-solving learning is difficult 1985; Barrows & Tamblyn, 1980) focus on prob-
and suggest only a brief template for applying lem-solving outcomes. They recommend
the events of instruction in the same way they instructional strategies, such as authentic cases,
treat concept-learning and ruleqearning out- simulations, modeling, coaching, and scaffold-
comes. The only instructional-design text that ing, to support their implicit problem-solving
systematically addresses problem solving outcomes, but they inadequately analyze or
(despite not referring to it as problem solving) is explicate the nature of the problems to be solved
the innovative text by van Merri~nboer (1997). It (Jonassen, 2000a).
focused on training the complex cognitive skills Jonassen (1997) began to describe the range
that are required to solve problems and uses dif- of problem-solving learning outcomes by distin-
ferent analysis processes that are based on tradi- guishing between well-structured and ill-struc-
tional, hierarchical task decomposition (which is tured problems in terms of their instructional
insufficient, some researchers believe, for ana- design requirements. Instructional designs for
lyzing the range of problem-solving outcomes; well-structured problems are rooted in informa-
see Jonassen, Tessmer, & Hannum, 1999 for tion-processing theory, but instructional designs
descriptions of alternative methods). Van for ill-structured problems necessarily share
Merri6nboer treated all problems the same. Yet assumptions with constructivism and situated
the most pervasive assumption of instructional cognition. Information-processing theories con-
design is that different learning outcomes neces- ceive of learning outcomes as generalizable
sitate different conditions of learning (GagnG skills that can be applied across content
1980). So, instruction to support problem-solv- domains, while constructivism and situated cog-
ing learning outcomes should differ from that nition argue for the domain specificity of any
used to support, for instance, concept learning performance and therefore recommend embed-
or rule learning. However, implied in compo- ding instruction in some authentic context
nent models of instructional design (e.g., Merrill, (Jonassen & Land, 2000). Assuming that differ-
Dick & Carey, Gagn4 & Briggs) is the belief that ent kinds of problem solving in different con-
identifying and learning the component con- texts and domains call on different skills,
cepts, rules, and principles that comprise a prob- Jonassen (1997) articulated instructional design
lem space enables learners to solve a problem. models for well-structured and ill-structured
Unfortunately "mastering each component skill problems. However, cognitive task analysis of
is not enough to promote nonroutine problem hundreds of problems has proven that this
solving" (Mayer, 1998, p. 50). If problem solving dichotomy is inadequate to accommodate the
is to be regarded as a separate type of learning or range and complexity of problem-solving out-
intellectual outcome, this assumption is prob- comes. Therefore, this article represents the next
lematic. An underlying assumption of this paper step in explicating the dimensions of problem
is that problems are not the same and so cannot solving in order to develop task-specific models
be supported in the same way as component for supporting the learning of problem solving. I
skills. Assuming that problem solving requires begin by defining the nature and characteristics
more than the acquisition of prerequisite skills, of problems and problem solving and conclude
specific models of problem-solving instruction by positing a typology of problem-solving out-
need to be proposed and tested. comes. Although some believe this endeavor to
Another reason for focusing on problem solv- be unachievable (Kahney, 1993), articulating cat-
ing is that it is at the center of practice according egories of problem solving is essential from an
to contemporary learning theories. Contempo- instructional design perspective, if we believe
TOWARD A DESIGNTHEORYOF PROBLEMSOLVING 65

that different learning outcomes necessitate differ- the problem space. Thinking is internalized
ent forms of ins~action. Subsequent papers will activity (Jonassen, 2000b). Conscious meaning
more completely articulate the cognitive, social, making is engaged by activity, so there is a
affective, and historical dimensions of this typol- reciprocal regulatory feedback between knowl-
ogy of problems and recommend instructional- edge and activity (Fishbein, Eckart, Lauver, van
design models for supporting learning of those Leeuwen, & Langemeyer, 1990). Problem solv-
problem-solving processes. ing requires manipulation of the problem space,
be it an internal mental representation or an
external physical representation.
W h a t Is a P r o b l e m ?
Over the past three decades, a number of
information-processing models of problem solv-
Just what is problem? There are only two critical
ing, such as the classic General Problem Solver
attributes of a problem. First, a problem is an
(Newell & Simon, 1972), have explained prob-
unknown entity in some situation (the difference
lem-solving processes. The General Problem
between a goal state and a current state). Those
Solver specifies two sets of thinking processes
situations vary from algorithmic math problems
associated with the problem-solving processes,
to vexing and complex social problems, such as
(a) understanding processes and (b) search pro-
violence in the schools. Second, finding or solv-
cesses. Another popular problem-solving
ing for the unknown must have some social, cul-
model, the IDEAL problem solver (Bransford &
tural, or intellectual value. That is, someone
Stein, 1984), describes problem solving as a uni-
believes that it is worth finding the unknown. If
form process of Identifying potential problems,
no one perceives an unknown or a need to deter-
Defining and representing the problem, Explor-
mine an unknown, there is no perceived prob-
ing possible strategies, Acting on those strate-
lem (whether the problem exists independent of
gies, and Looking back and evaluating the
any perception is an ontological issue that is
effects of those activities. Although the IDEAL
beyond the scope of this paper). Finding the
model assumes that these processes are applied
unknown is the process of problem solving.
differently to different problems, no explicit sug-
Problem solving is "any goal-directed gestions are made for how to do this. Gick (1986)
sequence of cognitive operations" (Anderson, synthesized these and other problem-solving
1980, p. 257). Those operations have two critical models (Greeno, 1978) into a simplified model of
attributes. First, problem solving requires the the problem-solving process, including the pro-
mental representation of the situation in the cesses of constructing a problem representation,
world. That is, human problem solvers construct
searching for solutions, and implementing and
a mental representation (or mental model) of the
monitoring solutions. Although descriptively
problem, known as the problem space (Newell
useful, these problem-solving models tend to
& Simon, 1972). Although there is little agree-
treat all problems the same in an effort to articu-
ment on the meanings of mental models or prob-
late a generalizable problem-solving procedure.
lem spaces, internal mental models (as opposed
The culmination of information-processing con-
to social or team mental models) of problem
cepts was an attempt to articulate a uniform the-
spaces are multimodal representations consist-
ory of problem solving (Smith, 1991), albeit
ing of structural knowledge, procedural knowl-
unsuccessfully.
edge, reflective knowledge, images and
metaphors of the system, and executive or stra- Problem solving is not a uniform activity.
tegic knowledge (Jonassen & Henning, 1999). Problems are not equivalent, in content, form, or
Although internal problem spaces may be exter- process. Schema-theoretic conceptions of prob-
nalized as formal models or representations lem solving opened the door for different prob-
using a variety of knowledge representation lem types by arguing that problem-solving skill
tools (Jonassen, 2000c), it is the mental construc- is dependent on a schema for solving particular
tion of the problem space that is the most critical types of problems. If the learner possesses a
for problem solving. Second, problem solving complete schema for any problem type, then
requires some activity-based manipulation of constructing the problem representation
66 ETR&D, Vol. 48, No. 4

Figure 1 [ ] Problem-solving skills.

Problem Variations ~ Representation Individual Differences = ProbIem Solving Skill

Ill-structuredness Context Domain knowledge
Complexity social familiarity
Abstractness/ historical perplexity
situatedness cultural experience
(domain specificity) Cues/Clues Structural knowledge
Modality Procedural knowledge
Systemic/conceptual knowledge
Domain-specificreasoning
Cognitivestyles
General problem-solving
strategies
Self-confidence
Motivation/perseverence

involves mapping an existing problem schema PROBLEM VARIATIONS

onto a problem and using the procedure that is
part of the problem schema to solve it. Existing Problems vary in their nature, in the way they
problem schemas are the result of previous are presented or represented, and in their com-
experiences in solving particular types of prob- ponents and interactions among them. Mayer
lems, enabling learners to proceed directly to the and Wittrock (1996) described problems as ill-
implementation stage of problem solving (Gick, defined-well-defined and routine-nonroutine.
1986) and try out the activated solution. Experts Jonassen (1997) distinguished weU-strnctured
a r e good problem solvers because they recog- from ill-structured problems and articulated dif-
nize different problem states that invoke certain ferences in cognitive processing engaged by
solutions (Sweller, 1988). If the type of problem each. Smith (1991) distinguished external fac-
is recognized, then little searching through the tors, including domain and complexity, from
problem space is required. Novices, who do not internal characteristics of the problem solver.
possess well-developed problem schemas, are There is increasing agreement that problems
not able to recognize problem types, so they vary in substance, structure, and process. In this
must rely on general problem solving strategies, section, I briefly describe the ways in which
such as the information processing approaches, problems vary. Problems vary in terms of their
which provide weak strategies for problem solu- structuredness, complexity, and abstractness
tions (Mayer, 1992). (domain specificity). Although there is similar-
In this paper, I attempt to articulate the attri- ity among these three factors, they are neither
butes of problems that make them different as independent nor equivalent. There is sufficient
well as some of the attributes of problem solvers independence among the factors to warrant sep-
that discriminate their abilities and dispositions arate consideration.
to solve problems. As depicted in Figure 1, I
believe that the ability to solve problems is a
function of the nature of the problem, the way Structuredness
that the problem is represented to the solver,
and a host of individual differences that mediate Jonassen (1997) distinguished well-structured
the process. Each of these factors will be from ill-structured problems and recommended
addressed in turn, although an elaborate discus- different design models for each, because each
sion of these factors is beyond the scope of a sin- kind of problem calls on different skills. The
gle paper. most commonly encountered problems, espe-
TOWARD A DESIGN THEORY OF PROBLEMSOLVING 67

cially in schools and universities, are well-struc- about the problem, so ill-structured problems
tured problems. Typically found at the end of are uniquely human interpersonal activities
textbook chapters and on examinations, these (Meacham & Emont, 1989).
well-structured problems require the applica-
The real problem-solving activity involved
tion of a finite number of concepts, rules, and with solving ill-structured problems is provid-
principles being studied to a constrained prob- ing a problem with structure when there is none
lem situation. These problems have been apparent (Simon, 1973).
referred to as transformation problems (Greeno,
I do not mean to imply that everyday practice
1978), which consist of a well-defined initial
does not include weU-structured problems. It
state (what is known), a known goal state
does (e.g., route planning). And we know that
(nature of the solution well defined), and a con-
ill-structured problems can become well-struc-
strained set of logical operators (known proce-
tured with practice (Simon, 1973). However,
dure for solving). Well-structured problems:
everyday practice is more suffused with ill-
• Present all elements of the problem to the structured problems than is educational prac-
learners. tice.
• Require the application of a limited number Researchers have long assumed that learning
of regular and well-structured rules and prin- to solve well-structured problems transfers pos-
ciples that are organized in predictive and itively to learning to solve ill-structured prob-
prescriptive ways. lems. Although information processing theories
believed that "in general, the processes used to
• Have knowable, comprehensible solutions
solve ill-structured problems are the same as
where the relationship between decision
those used to solve well structured problems"
choices and all problem states is known or
(Simon, 1978, p. 287), more recent research in sit-
probabilistic (Wood, 1983).
uated and everyday problem solving makes
Ill-structured problems, on the other hand, clear distinctions between thinking required to
are the kinds of problems that are encountered solve well-structured problems and everyday
more often in everyday and professional prac- problems. Dunkle, Schraw, and Bendixen (1995)
tice, so they are typically emergent. Because they concluded that performance in solving well-
are not constrained by the content domains defined problems is independent of perfor-
being studied in classrooms, their solutions are mance on ill-defined tasks, with ill-defined
not predictable or convergent. Ill-structured problems engaging a different set of epistemic
problems may also require the integration of beliefs. Hong, Jonassen, and McGee (in press)
several content domains. For example, solutions found that solving ill-structured problems in a
to problems such as pollution may require the simulation called on different skills than solving
application of concepts and principles from well-struc~red problems, including metacogni-
math, science, political science, and psychology. tion and argumentation. Jonassen and Kwon (in
Ill-structured problems appear ill-structured press) showed that communication patterns in
because they: teams differed when solving well-structured
• Possess problem elements that are unknown and ill-structured problems. Clearly more
or not known with any degree of confidence research is needed to expand these findings, yet
(Wood, 1983). it seems reasonable to predict that well-struc-
tured and ill-structured problem solving engage
• Possess multiple solutions, solution paths, or different intellectual skills.
no solutions at all (Kitchner, 1983).
• Possess multiple criteria for evaluating solu-
tions, so there is uncertainty about which Complexity
concepts, rules, and principles are necessary
for the solution and how they are organized. Problem complexity is defined by the number of
• Often require learners to make judgments issues, functions, or variables involved in the
and express personal opinions or beliefs problem; the degree of connectivity among
68 ETR&D,Vol,48, No. 4

those properties; the type of functional relation- Domain Specificity (Abstract-Situated)

ships among those properties; and the stability
among the properties of the problem over time Contemporary research and theory in problem
(Funke, 1991). Although Funke also addressed solving claims that problem-solving skills are
the availal~ity of information and contextual domain-and-context-specific. That is, problem-
richness (factors more related to structuredness), solving activities are situated, embedded, a n d
complexity is more concerned with how many, therefore dependent on the nature of the context
how clearly, and how reliably components are or domain. This is because solving problems
represented implicitly or explicitly in the prob- within a domain relies on cognitive operations
lem. The most complex problems are dynamic, that are specific to that domain (Mayer, 1992;
that is, those in which the task environment and Smith, 1991; Sternberg & Frensch, 1991). These
its factors change over time. are often referred to as strong methods, as
opposed to domain-general strategies (weak
Problem difficulty is a function of problem methods). For example, Lehman, Lempert, and
complexity. For example, problem difficulty has Nisbett (1988) concluded that different forms of
been found to be a function of relational com- reasoning are learned in different graduate dis-
plexity (English, 1998). The idea of problem ciplines. Graduate students in the probabilistic
complexity seems to be intuitively recognizable sciences of psychology and medicine perform
by even untrained learners (Suedfield, de Vries, better on statistical, methodological, and condi-
Bluck, & Wallbaum, 1996). Problem complexity tional reasoning problems than do students in
necessarily affects learners' abilities to solve law and chemistry, who do not learn such forms
problems. For example, problem complexity has of reasoning. The cognitive operations are
significant effects on search problems (Halgren learned through the development of pragmatic
& Cooke, 1993). reasoning schemas rather than exercises in for-
mal logic. Graduates in different domains
Why do we assume that complex problems
develop reasoning skills through solving situ-
are more difficult to solve than simple prob-
ated, ill-structured problems that require forms
lems? The primary reason is that complex prob-
of logic that are domain-specific.
lems involve more cognitive operations than
Ill-structured problems tend to be more situ-
simpler ones (Kluwe, 1995). Therefore, working
ated, but well-structured problems tend to rely
memory requirements increase at least propor-
more on general problem-solving skills, such as
tionally. Accommodating multiple factors dur-
means-ends analysis. However, well-structured
ing problem structuring and solution generation
problems, in the form of story problems, can be
places a heavy burden on working memory. The
quite situated while ill-structured problems, in
more complex a problem, the more difficult it
the form of dilemmas, can be fairly abstract.
will be for the problem solver to actively process
the components of the problem.
Complexity and structuredness overlap, hi- PROBLEM REPRESENTATIONS
structured problems tend to be more complex,
especially those emerging from everyday prac- Problems also vary in terms of how they are rep-
tice. Most well-structured problems, such as resented to and perceived by the problem solver.
textbook math and science problems, tend to Problems in everyday and professional contexts
engage a constrained set of variables that behave are embedded in those contexts, which requires
in predictable ways. Although ill-structured the problem solver to disambiguate important
problems tend to be more complex, well-struc- from irrelevant information in the context and
tured problems can be extremely complex and construct a problem space that includes relevant
ill-structured problems fairly simple. For exam- information from the context. Experts within a
ple, video games can be very complex, well- domain often use artificial symbol systems that
structured problems, while selecting what to are specific to the domain for representing prob-
wear from our closets for different occasions (at lems (Goel & Pirotli, 1989). For example, physi-
least for me) is a simple ill-structured problem. cians and computer programmers are often
TOWARDA DESIGNTHEORYOF PROBLEMSOLVING 69

criticized for communicating to their clients in ety, and stress also affect problem solving, as
inexplicable code. In addition to making prob- they do most performance, in a curvilinear rela-
lems harder to represent to novices, these sym- tionship (i.e., the U-shaped arousal curve). Space
bol systems also insure the domain-specificity of does not allow consideration of those here.
the problem representations and solutions.
An important function of designing for prob- Familiarity
lem solving is deciding how to represent the
problem to novice learners. Problems that are Perhaps the strongest predictor of problem-solv-
represented to learners in formal learning situa- ing ability is the solver's familiarity with the
tions (e.g., schools, universities, and training) problem type. Experienced problem solvers
are usually simulations of everyday and profes- have better developed problem schemas, which
sional problems, so instructional designers can be employed more automatically (Sweller,
decide which problem components to include 1988). Although familiarity with a type of prob-
and how to represent them to the learner. That lem facilitates problem solving, that skill seldom
is, designers assume responsibility for construct- transfers to other kinds of problems or even to
ing the problem space for the learners. In order the same kind of problem represented in
to do so, designers provide or withhold contex- another way (Gick & Holyoak, 1980,1983).
tual cues, prompts, or other clues about informa- Mayer and Wittrock (1996) refer to routine
tion that need to be included in the learner's and nonroutine aspects of problems. Routine
problem space. How consistent and overt those problems are obviously familiar to learners and
cues are will determine problem difficulty and consequently more transferable (at least within a
complexity. Additionally, designers make deci- task environment). Therefore routine problems
sions about the modality and medium for repre- appear more well-structured to the experienced
senting different problem components. solver. Transfer of nonroutine problems (those
An important issue in problem representa- not familiar to the problem solver) require high-
tion is the fidelity of the representation. Is the road transfer (far transfer), which is more effort-
problem represented in its natural complexity fial and conscious whereas routine problems rely
and modality, or is it filtered when simulated? more on low-road transfer (near transfer), which
Should social pressures and time constraints be involves less conscious attention (Salomon &
represented faithfully? That is, does the problem Perkins, 1989).
have to be solved in real time, or can it be solved
in leisure time? What levels of cooperation or
D o m a i n a n d Structural K n o w l e d g e
competition are represented in the problem?
These are but a few of the decisions that design-
Another strong predictor of problem-solving
ers must make when representing problems for
skills is the solver's level of domain knowledge.
learning.
How much someone knows about a domain is
important to understanding the problem and
INDIVIDUAL DIFFERENCES
generating solutions. However, that domain
knowledge must be well integrated in order to
Smith (1991) distinguished between internal and support problem solving. The integratedness of
external factors in problem solving. External fac- domain knowledge is best described as struc-
tors are the variations in problem type and rep- tural knowledge (Jonassen, Beissner, & Yacci,
resentation, as just described. Internal factors are 1993). Structural knowledge is the knowledge of
those that describe variations in the problem how concepts within a domain are interrelated.
solvers. Just as individual differences mediate It is also known as cognitive structure, the orga-
other kinds of learning, they also mediate learn- nization of relationships among concepts in
ing how to solve different kinds of problems. memory (Shavelson, 1972).
Some of the trait characteristics of learners that Domain knowledge and skills are very
may affect problem solving are described next. important in problem solving. Structural knowl-
Other state characteristics, such as fatigue, anxi- edge may be a stronger predictor of problem
70 ErR&D,VoL48, No. 4

solving than familiarity. Robertson (1990) found standing, the use of information to achieve a
that the extent to which think-aloud protocols goal, and the assessment of learning progress.
contained relevant structural knowledge was a Because of the cognitive and affective demands
stronger predictor of how well learners would of problem solving, the role of metacognition
solve transfer problems in physics than either has received considerable research attention. In
attitude or previous experience solving similar fact, metacognifive actions are regarded as a
problems. Structural knowledge that connects driving force in problem solving along with
formulas and important concepts in the knowl- beliefs and attitudes (Lester, 1994). When solv-
edge base are important to understanding phys- ing mathematics problems, good problem solv-
ics principles. Gordon and Gill (1989) found that ers work to clarify their goals, understand the
the similarity of learners' graphs (reflective of concepts and relationships among the elements
underlying cognitive structure) to those of of a problem, monitor their understanding, and
experts was highly predictive of total problem- choose and evaluate actions that lead toward the
solving scores (accounting for over 80% of the goal (Gourgey, 1998). Problem solving requires
variance) as well as specific problem-solving knowing not only what to monitor but also how
activities. Well-integrated domain knowledge is to monitor one's performance and sometimes
essential to problem solving. unlearning bad habits (Lester, 1994). The devel-
opment of metacognifive skills enables students
to strategically encode the nature of the problem
Cognitive Controls by forming mental representations of the prob-
lems, select appropriate plans for solving the
Individuals also vary in their cognitive styles
problem, and identify and overcome obstacles to
and controls, which represent patterns of think-
the process (Davidson & Sternberg, 1998). Ori-
ing that control the ways that individuals pro- enting and self-judging are important
cess and reason about information (Jonassen & metacognitive skills that are positively related to
Grabowski, 1993). Cognitive controls, such as
problem-solving performance, and they can be
field independence, cognitive complexity, cogni-
learned (Masui & DeCorte, 1999).
tive flexibility, and category width are most
likely to interact with problem solving. Some Like most other research issues, research on
research has supported this belief. For example, the role of metacognition in problem solving has
field independents are better problem solvers focused primarily on solving mathematical
than field dependents (Davis & Haueisen, 1976; story problems, which are typically well-struc-
Heller, 1982; Maloney, 1981; Ronning, tured. Although Hong, Jonassen, and McGee (in
McCurdy, & Ballinger, 1984) because of their press) found that the application of metacogni-
ability to attend to salient cues. Learners with rive skills is more important to solving ill-struc-
higher cognitive flexibility and cognitive com- tured problems than weU-structured problems,
plexity should be better problem solvers than almost no research on the role of metacognition
cognitive simplistic learners because they con- in solving ill-structured problems exists. There is
sider more alternatives (Stewin & Anderson, no doubt that metacognition and self-regulation
1974) and they are more analytical. Although the of cognitive performance are essential compo-
relationship between cognitive controls and nents of all types of problems, although it is
problem solving needs to be better examined, it likely that the specific requisite skills will vary
is reasonable to predict that learners who think with problem type.
in ways that are more analytical should be better
problem solvers.
Epistemological Beliefs

Metacognition Problem solving, especially ill-structured prob-

lem solving, often requires solvers to consider
Flavell (1979) described metacognition as the the veracity of ideas and multiple perspectives
awareness of how one learns, the ability to judge while evaluating problems or solutions. The
the difficulty of a task, the monitoring of under- ability to do so depends partially on their under-
TOWARDA DESIGNTHEORYOF PROBLEMSOLVING 71

lying beliefs about knowledge and how it devel- However, cognitive processes are necessary but
ops. That is, learners' epistemic beliefs about the insufficient requirements for solving problems,
nature of problem solving also affect the ways especially complex and ill-structured ones. They
that they naturally tend to approach problems. require significant affective and conative ele-
A number of epistemological theories have been ments as well (Jonassen & Tessmer, 1996).
related to a broad range of learning outcomes
Affective elements, such as attitudes and
(Hofer & Pintrich, 1997).
beliefs about the problems, problem domain,
The best-known theory of epistemic beliefs
and the learner's abilities to solve the problem,
was developed by William Perry (1970). He dis-
significantly affect a problem solver's abilities. If
tinguished nine separate stages of intellectual
development clustered into three periods. In the problem solvers are predisposed to certain prob-
first period, dualist learners believe that knowl- lem solutions because of personal beliefs, then
edge is right or wrong, that teachers and profes- they will be less effective because they overrely
sors have the right knowledge, and that the role on that solution.
of students is to assimilate what the teacher Conative (motivational and volitional) ele-
knows. Their absolutist beliefs stress facts and
ments, such as engaging intentionally, exerting
truth. In the second period, multiplicity repre-
effort, persist~g on task, and making choices,
sents the acceptance of different perspectives
also affect the effort that learners will make in
and skepticism about expertise in general.
Multiplists rely on methods and processes to trying to solve a problem. Knowing how to solve
establish truth. In the third period, contextual problems, believing that you know how to solve
relativistic, evaluative thinkers accept the role of problems, and exerting the effort to do so are
judgment and wisdom in accommoda~ng often dissonant. Students think harder and pro-
uncertainty, and that experts may provide better cess material more deeply when they are inter-
answers. However, ideas must be evaluated for ested and believe that they are able to solve the
their merits and the cultural and intellectual per- problem (i.e., have high self-efficacy), according
spectives from which they derive. to Mayer's (1998) effort-based learning princi-
More complex and ill-structured problems ple. Problem solving requires a number of affec-
require higher levels of epistemic belief, which tive dispositions, especially self-confidence, and
most students have not yet developed. Unfortu- beliefs and biases about the knowledge domain.
nately, as a result of the preponderance of algo- For example, Perkins, Hancock, Hobbs, Martin,
rithmic teaching approaches in mathematics, for and Simmons (1986) found that some students,
instance, there is "a belief by students that math-
when faced with a computer programming
ematical problems are solved by applying proce-
problem, would disengage immediately, believ-
dures that a person may or may not know"
ing that it was too difficult, while others would
(Greeno, 1991, p. 83). There is a right and wrong
way to do things. Solving more complex and ill- keep trying to find a solution. If problem solvers
structured problems depends on multiplicitous do not believe in their ability to solve problems,
and contextual relativistic thinking. Although they will most likely not exert sufficient cogni-
no research has connected epistemic beliefs and tive effort and therefore not succeed. Their self-
problem solving, the relationship is obvious and confidence of ability will predict the level of
needs to be examined, especially in ill-struc- mindful effort and perseverance that they will
tured problem solving. apply to solving the problem. Greeno (1991)
claimed that most students believe that if math
problems have not been solved in a few minutes,
Affectlve and Conative
the problem is probably unsolvable and there is
Mayer (1992) claimed that the essential charac- no point in continuing to try, despite the fact that
teristics of problem solving are directed cogni- skilled mathematicians often work for hours on
tive processing. Clearly, problem solving a problem. Task persistence and effort are strong
requires cognitive and metacognitive processes. predictors of problem-solving success.
72 E'fR&D, Vol. 48, No. 4

General Problem-Solving Skills 2000). Some of these problem classes have been
extensively researched, such as logical prob-
There is a general belief that some people are lems, story (word) problems, and decision-mak-
better problem solvers than others because they ing problems, as evidenced by significant
use more effective problem-solving strategies. literature bases. Others, such as rule-using and
That depends on the kind of strategies they use. strategic performance, represent new classes.
Solvers who attempt to use weak strategies, such
The typology of problems in Table 1 repre-
as general heuristics like means-ends analysis
sents a continuum of problems from left to right
that can be applied across domains, generally
as well-structured to ill-structured (Simon
fair no better than those who do not (Singley &
(1973). Also, to some undetermined extent, the
Anderson, 1989). However, solvers who use
typology is taxonomic, with the well-structured
domain-specific, strong strategies are better
problems on the left as prerequisite to the ill-
problem solvers. Experts effectively use strong
structured problems on the right. Case-analysis
strategies, and some research has shown that
problems, for instance, require problem solvers
less experienced solvers can also learn to use
to be able to solve decision making and aspects
them (Mayer & Wittrock, 1996).
of troubleshooting (hypothesis generation and
testing) in order to be able to solve case prob-
Summary lems. Decision making requires rule using and
story problems as prerequisite, and so on.
Problem solving varies along at least three dif-
ferent dimensions: (a) problem type, (b) problem This typology assumes that there are similar-
representation, and (c) individual differences. ities in the cognitive processes required to solve
As described earlier, problems vary in terms of each type of problem. It also assumes that some
structuredness, complexity, and abstractness. instructional strategies can be generalized across
Problem representations vary by context and problem types. Though this assumption contra-
modality. A host of individual differences medi- dicts the domain specificity principle that domi-
ate individuals" abilities to solve those problems. nates current theory, it is not clear yet how many
Although dichotomous descriptions of general or how completely these instructional strategies
types of problems, such as well-structured and can be generalized.
ill-structured, are useful for clarifying attributes It is also important to note that these problem
of problems, they are insufficient for isolating classes are neither absolute nor discrete. Addi-
required cognitive and affective processes and tional analysis of hundreds or even thousands of
suggesting appropriate instructional strategies problems is needed. Additional research may
for supporting how to learn to solve problems. possibly identify new categories or reorganize
Additional clarity is needed to resolve specific the existing categories. Likewise, these classes
problem types. are not discrete or independent of each other,
that is, they are not mutually exclusive catego-
ries. So, there are necessarily similarities and
TYPOLOGY OF PROBLEM SOWING overlap among the classes. Additionally, the
true nature of a problem will depend on the
In order to articulate different problem types problem solver's experience and the nature of
(shown in Table 1), hundreds of problems (see the context in which the problem is encountered.
Table 2 for examples) were collected. A cogni- A similar problem in different contexts or with
tive task analysis of those problems was con- different learners may assume the characteris-
ducted in order to identify attributes of those tics of a different class of problem.
problems. Next, an iterative sort of those prob- Table I lists horizontally 11 different types of
lems, based on their characteristics, distin- problem-solving outcome, including (a) logical,
guished 11 different types of problems (see (b) algorithmic, (c) story, (d) rule-using, (e) deci-
Table I for a description of problem types, Table sion making, (0 troubleshooting, (g) diagnosis-
2 for examples; hundreds of sample problems solution, (h) strategic performance, (i) case
will be archived in a Web site by the end of analysis, (j) design, and (k) dilemma. Within
TOWARD A DESIGN THEORY OF PROBLEMSOLVING 73

each category, problems vary with regard to lems to formally identical problems (Hayes &
abstractness and complexity. The specific learn- Simon, 1977; Reed, Ernst, & Banerji, 1974).
ing outcome for each of these problem types is Logical problems can be decidedly more
described in the next row, followed by the complex than these. However, few if any logical
inputs to the problem-solving process. The next problems are embedded in any authentic con-
row describes criteria for judging the success of text, making them necessarily more abstract and
problem solutions. Well-structured problems therefore less transferable. Logical problems
focus on correct, efficient solutions, while the ill- have been the focus of considerable laboratory-
structured problems focus more on decision based psychological research. However, the use-
articulation and argumentation. Problems vary fulness of that research to instructional design is
from logic problems and algorithms with exact limited by the lack of ecological validity.
solutions to dilemmas with no verifiably correct
solution. The role of problem context is listed
next. The role of context becomes vitally impor- Algorithmic Problems
tant in defining ill-structured problems, while
well-structured problems de-emphasize the role One of the most common problem types
of context. Finally, Table 1 describes the encountered in schools is the algorithm. Most
structuredness and abstractness (described ear- common in mathematics courses, students are
lier) that are typical of that class of problems. taught to solve problems such as long division
Complexity is not included because it varies too or equation factoring using a finite and rigid set
much within problem class to describe consis- of procedures with limited, predictive decisions.
tently. Table 1 presents a brief overview of the Solving algorithms requires number compre-
different kinds of problems that practitioners hension, number production, and calculation
and learners need to learn to solve. A number of (McCloskey, Caramaza, & Basili, 1985).
examples of each problem type is listed in Table Learners' number-processing systems, com-
2. In this section of the paper, I briefly describe prised of comprehending and producing num-
each kind of problem solving. In the next paper bers, are the conceptual understandings that
in this series, I provide a cognitive model of the complement the calculation procedures. Calcu-
processes involved in solving each kind of prob- lation, according to McCloskey et al. (1985),
lem. These processes are based on a cognitive requires comprehension of the operations (e.g.,
task analysis, but need to be validated and fur- associative and commutative properties and
ther explicated by observation, interviewing, concepts of multiplication and division), execu-
and artifact analysis of problem solutions. tion procedures for calculating, and retrieval of
arithmetic facts (e.g., times tables). Such algo-
rithmic approaches are also commonly used in
Logical Problems science or home economics courses. Most reci-
pes are algorithms for cooking. It is likely that a
model similar to that proposed by McCloskey et
Logical problems tend to be abstract tests of rea-
al. can be generated for nonmathematical forms
soning that puzzle the learner. They are used to
of algorithmic problems.
assess mental acuity, clarity, and logical reason-
ing. Classic games such as missionaries and can- Many researchers, such as Smith (1991),
nibals or Tower of Hanoi challenge learners to argue that algorithms (repeating a series of
find the most efficient (least number of moves) steps) are, by nature, not problems. When learn-
sequence of action. Rubic's Cube TM was a popu- ers are required to select and perhaps modify an
lar game in the 1970s requiring the user to rotate algorithm for use in an exercise, it may become
the rows and columns of a three-dimensional problem solving. However, because algorithms
cube to form patterns. In each of these, there is a are so generally considered to represent prob-
specific method of reasoning that will yield the lems, for better or worse, they should be
most efficient solution. The learner is required to included in the typology.
discover that method. Research has shown that The primary limitation of algorithmic
there is no transfer of solutions of these prob- approaches is the overreliance on procedurally
74 ETR&D,Vol.48, No. 4

Table 1 [ ] A description of problem types,

Rule- Decision
Logical Algorithmic Story Using making
Problems Problems Problems Problems Problems

Learning logical procedural disambiguate procedural identifying

Activity control and sequence of variables; process benefits and
manipulation manipulations; select and constrained limitations;
of limited algorithmic apply by rules; weighting
variables; process algorithm to select and options;
solve puzzle applied to produce apply rules selecting
similar sets correct to produce alternative
of variables; answer using system- and justifying
Calculating prescribed constrained
or producing method answers or
correct answer products

Inputs puzzle formula or story with situation in decision

procedure formula or constrained situation with
procedure system; limited
embedded finite rules alternative
outcomes

Success Criteria efficient answer or answer or productivity answer or

manipulation; product matches product matches (number of product
number of in values in values and relevant or matches
moves or and form form; correct useful in values
manipulations algorithm answers or and form
required used products

Context abstract task abstract, constrained to purposeful life decisions

formulaic pre-defined academic,
elements, real world,
shallow context constrained

Stucturedness discovered procedural well-defined unpredicted finite

predictable problem classes; outcome outcomes
procedural
predictable

Abstractness abstract, abstract, limited need-based personally

discovery procedural simulation situated
Tablecan~nu~

oriented k n o w l e d g e structures and the lack or of conceptual understanding of the underlying

absence of conceptual understanding of the processes. This is a common complaint about
objects of the algorithm and the procedures learning statistics, w h e r e professors focus on the
engaged. Content that is learned only as a proce- algorithms and miss the p u r p o s e of studying the
d u r e can rarely be transferred because of a tack statistical analysis. Learners w h o are a d e p t at
TOWARD A DESIGNTHEORYOF PROBLEMSOLVING 75

Table 1 [ ] Continued.

Trouble- Diagnosis- Strategic Case

shooting Solution Perfvrmance Analysis Design
Problems Problems Problems Problems Problems Dilemmas
examine troubleshoot applying solution acting on reconciling
system; run system faults; tactics identification, goals to complex,
tests; evaluate select and to meet alternative produce non-
results; hypo- evaluate strategy in actions, artifact; predictive,
thesize and treatment real-time, argue problem vexing
confirm fault options and complex position structuring decision
states using monitor; performance & with no
strategies (re- apply maintaining articulation solution;
place, serial problem situational perspectives
elimination, schemas awareness irreconcil-
space split) able

malfunctioning complex real-time, complex, vague goal situation

system with one system with complex leisure-time statement with
or more faults faults and performance system with with few antinomous
numerous with competing multiple constraints; positions
optional needs ill-defined requires
solutions goals structuring

fault(s) strategy used; achieving multiple, multiple, articulated

identification; effectiveness strategic unclear undefined preference
efficiency of and efficiency objective criteria; no with some
fault isolation; of treatment; right or wrong justification
justification of ---only better
treatment or worse
selected

closed system real world, real-lime real world, complex, topical,

real world technical, performance constrained real world; complex,
mostly closed degrees of inter-
system freedom; disciplinary
limited input
& feedback

finite faults & finite faults & ill-structured iU-structured ill-structured finite
outcomes outcomes strategies; outcomes,
well-structured multiple
tactics reasoning

problem problem contxtually case problem issue

situated situated situated situated situated situated

abstract reasoning can learn increasingly com- Ston/Problems

plex algorithms, such as those encountered in
calculus, trigonometry, and other mathematics In an attempt to situate algorithms in some k i n d
domains. Global reasoning learners are limited of context, m a n y textbook authors and teachers
in their ability to create such abstract representa- e m p l o y story problems (also k n o w n as w o r d
tions of procedures, so they encounter problems. problems). Story p r o b l e m s have been exten-
76 ETR&D,Vol.48, No. 4

Table 2 [ ] Examples of problem types.

LogicalProblems
Tower of Hanoi; Cannabals & Missionaries; how can I divide the water in the first jug and second jug using
only three jugs; Rubic's Cube; draw four straight lines on 3 x 3 array of dots without removing pen from
paper; divide triangular cake into four equal pieces

Algorithms
Factor quadratic equation; convert Farenheit to Celsius temperatures; bisect any given angle

Story Problems
How long for car A to overtake car B traveling at different speeds; apply Boyle's law to problem statement;
calculate reagents needed to form a specific precipitate in a chemical reaction; most back-of-the-chapter
textbook problems

Rule-Using Problem
Search an online catalog for best resources; expand recipes for 10 guests; how many flight hours are required
to pay off a 777; prove angles of isosoles tTiangle are equal; calculating material needed for addition; change
case to subjunctive

Decision-MakingProblems
Should I move in order to take another job; which school should my daughter attend; which benefits package
should I select; which strategy is appropriate for a chess board configuration; how am I going to pay this bill;
what's the best way to get to the interstate during rush hour; how long should my story be

TroubleshootingProblems
Troubleshoot inoperative modem; why won't car start; determine chemicals present in qualitative analysis;
determine why newspaper article is poorly written; identify communication breakdowns in a committee;
determine why local economy is inflationary despite national trends; isolate cause of inadequate elasticity in
polymer process; why are trusses showing premature stressing; why is milk production down on dairy farm

Diagnosis-Solution Problems
Virtually any kind of medical diagnosis and treatment; how should I study for the final exam; identifying
and treating turfgrass problems on a golf course; develop individual plan of instruction for special education
students

StrategicPerformance
Flying an airplane; driving a car in different conditions; managing investment portfolio; how can I avoid
interacting with person X; moving to next level in Pokemon game; teaching in live class; arguing points of
law before court

Situated Case-PolicyProblems
Harvard business cases; plan a menu for foreign dignitaries; render judgment in any tort case; develop policy
for condominium association; evaluate performance of a stock portfolio; how should Microsoft be split up

Design Problems
Design instructional intervention given situation; write a short story; compose a fugue; design a bridge; make
a paper airplane; design a dog house; design a vehicle that flies; developing curriculum for school; plan
marketing campaign for new Internet company; develop investment strategy for money market fund

Dilemmas
Should abortions be banned; resolve Kosovo crisis; negotiate peace between Hutus and Tutsis in Rwanda;
redistribute wealth through tax policies; develop bipartisan bill for U.S. Congress that will pass with 2/3
majority

Note: Conditions (parameters, constraints, restraints) and performance standards not included. These necessarily affect the
structuredness, complexity, and abstractness of problems and so will affect categorization of the problems.
TOWARDA DESIGNTHEORYOF PROBLEMSOLVING 77

sivety researched, especially in the domain of predicts that an otherwise boring task cannot be
mathematics. Mathematical values are embed- made interesting by adding a few interesting
ded in a brief story. Learners are required to details" (Mayer, 1998, p. 57).
identify key words in the story, select the appro-
priate algorithm and sequence for solving the
Rule-Using Problems
problem, and apply the algorithm. It is hoped
they also will check their responses (Sherrill, Many problems have correct solutions but mul-
1983). This is a more complex cognitive process tiple solution paths or multiple rules governing
than simply applying an algorithm, as described the process. They tend to have a clear purpose or
in the previous class. goal that is constrained but not restricted to a
When problems are more complex (that is, specific procedure or method. Rule-using prob-
consist of multiple steps), Sherrill found that lems can be as simple as expanding a recipe to
learners perform more poorly because they accommodate more guests and as complex as
attend to the wrong words and therefore apply completing tax return schedules. Using an onl-
the wrong algorithm. When learners focus too ine search system to locate a library's holdings
closely on surface features or recall familiar or using a search engine to find relevant infor-
solutions from previously solved problems, they mation on the World Wide Web are examples of
fail to transfer story problem skills to other prob- rule-using problems. The purpose is clear: find
lems (Woods et al., 1997). They fail to under- the most relevant information in the least
stand the principles and the conceptual amount of time. That requires selecting search
applications underlying the performance, so terms, constructing effective search arguments,
they are unable to transfer the ability to solve implementing the search strategy, and evaluat-
one kind of problem to problems with the same ing the utility and credibility of information
structure but dissimilar features. That is why found. Schacter, Chung, and Dorr (1998) found
many researchers have emphasized the cogni- that students rarely employ systematic search
tive representation of the information in the strategies and spend little to no time planning
story. their searches. This is the rule-oriented essence
of searching. Given that multiple search strate-
Through practice, learners construct schemas
gies are possible, rule-using problems can
for problems they solve. Marshall (1995) identi-
become decidedly more ill-structured.
fied five different story problem schemas
Popular card games such as bridge or hearts
(change, group, compare, restate, and vary) and
has found that when learners practice those and board games such as checkers and chess are
more complex forms of rule-using problems.
forms of problem classification, their perfor-
These games employ more complex rules and
mance improves dramatically. Solving story
constraints. Current computer games, such as
problems is more difficult and ill-structured
Pokemon TM, are also forms of complex rule-
than solving algorithms because it requires the
using problems.
semantic comprehension of relevant textual
information, the capacity to visualize the data, Rule-using problems constitutes a new dass
the capacity to recognize the deep structure of of problem solving, so no research about these
the problem, the capacity to correctly sequence specific kinds of problems exists. Cognitive pro-
the solution activities, and the capacity and will- cesses and design principles will have to be gen-
ingness to evaluate the procedure that was used eralized from any research on prototypic
to solve the problem (Lucangelli, Tressoldi, & examples of this class (e.g., online searching) in
Cendron, (1998). These skills become even more addition to cognitive task analysis.
important when solving college-level story
problems in physics and engineering, for Decision-Making Problems
instance. Notwithstanding these skills, solving
story problems remains a relatively simple and Decision-making problems typically involve
well-structured activity that students often do selecting a single option from a set of alterna-
not take seriously. Why? "Interest theory . . . tives based on a set of criteria. Decision makers
78 ETR&D,VoL 48, No. 4

must choose from a set of alternatives, each of domains is synonymous with problem solving,
which has one or more conseq':_~_nces. Mullen perhaps because the inoperative entities that
and Roth (1991) describe decision-making as a involve troubleshooting are most easily per-
process that includes recognizing problems and ceived as problems. Mechanics who trouble-
analyzing values (changing the present course shoot your inoperative car or computer
of one's life for opportuni',y or to avoid detri- programmers who debug your inoperative com-
ment); generating alternative choices (gathering puter are always recognized as problem solvers.
information about choices); evaluating choices The primary purpose of troubleshooting is fault
(identifying best choice [optimizing] or satisfy- state diagnosis. That is, some part or parts of a
ing some external criteria [satisficing] by analyz- system are not functioning properly, resulting in
ing cost and benefits of outcomes); and binding a set of symptoms that have to be diagnosed and
the will (committing to choice) and ignoring matched with the user's knowledge of various
sunk costs (effort already expended). Decision- fault states. Troubleshooters use symptoms to
making is an everyday part of life. In fact, many generate and test hypotheses about different
of the decisions listed in Table 2 are referred to fault states. In the spiral of data collection,
as life dilemmas, the decisions of which are most hypothesis generation, and testing, the trouble-
affected by riskiness ~orgas, 1982). People shooter refines hypotheses about the fault state.
avoid risk when the outcomes are positive and
Troubleshooting skill requires system knowl-
embrace risk when the outcomes are negative
edge (how the system works), procedural
(Tversky & Kahneman, 1981).
knowledge (how to perform problem-solving
The decision-making literature is divided procedures and test activities), and strategic
into normative decision theory, which present knowledge (strategies such as search-and-
models of how decisions ought to be made, and replace, serial elimination, and space splitting)
empirical decision theory, which describes how (Pokorny, Hall, Gallaway, & Dibble, 1996).
people actually make decisions (Mullen & Roth, These skills are integrated and organized by the
1991). Wrestling with decisions is not always a troubleshooter's experiences. The trouble-
rational process. Rather it is fraught with psy- shooter's mental model consists of conceptual,
chological impediments, induding conformity functional, and declarative knowledge, includ-
and social pressure, various forms of stress, cog- ing knowledge of system components and inter-
nitive dissonance, fear or failure, and many oth- actions, flow control, fault states (fault
ers. characteristics, symptoms, contextual informa-
Decision making varies in complexity. Simple tion, and probabilities of occurrence), and fault-
decisions with a single solution and a limited testing procedures. The primary differences
number of choices are more likely to be solved between expert and novice troubleshooters are
through some form of rational analysis. How- the amount and organization of system knowl-
ever, decision making in multistage, dynamic edge (Johnson, 1988). TroubleshootGng requires
task environments, where conditions are chang- an integrated understanding of how the system
ing, requires that the problem solver make multi- being troubleshot works, which is best taught
ple decisions under risk and uncertainty, and through functional flow diagrams (Johnson &
that those decisions be made in real time (e.g., Satchwell, 1993).
fighting a spreading fire or treating a medical
The most effective method for analyzing
patient with a deteriorating condition; Kerstholt
troubleshooting problems, the PARI (precursor,
& Raaijmakers, 1997). Dynamic decision making
action, result, interpretation) method, attempts
is an integral part of strategic performance prob-
lems (described later). to identify each Action (or decision) that the
problem solver performs, the Precursor (or Pre-
requisite) to that action, the Result of that action,
Troubleshooting Problems and an expert's Interpretation of the results of
the action. Experts are probed for the reasons
Troubleshooting is among the most common and assumptions behind their actions while they
forms of everyday problem solving and in many are solving problems. They are then asked to
TOWARDA DESIGNTHEORYOF PROBLEMSOLVING 79

elaborate on their solutions, focusing especially adjusted. Those adjustments are contextually
on reasoning that they use in making their deci- constrained. Strategic performances can be com-
sions about what to do (Hall, Gott, & Pokorny, plex. The options can be quite numerous and
1995). their implementation quite complex.

Strategic P e r f o r m a n c e Case-Analysis Problems

Strategic performance involves real-time, com- Case-analysis problems, ironically, emerge from
plex and integrated activity structures, where instruction, not from reality. The case method of
the performers use a number of tactics to meet a instruction emerged at Harvard Law School
more complex and ill-structured strategy while nearly 130 years ago (Williams, 1992). Analyz-
maintaining situational awareness. In order to ing cases, preparing briefs, and defending judg-
achieve the strategic objective, such as flying an ments are all authentic activities for law
airplane in a combat mission or quarterbacking students. In business and many other profes-
a professional football offense, the performer sional contexts, such as internal relations (Voss,
applies a set of complex tactics that are designed Wolfe, Lawrence, & Engle, 1991) and managerial
to meet strategic objectives. Strategy formation problem solving (Wagner, 1991), analyzing
represents a situated case or design problem complex, situated case problems defines the
(described next). Meeting that strategy through nature of work. Business problems, including
tactical maneuvers is a strategic performance. production planning, are common case prob-
The difficulty arises from the real-time decision lems. For example, case problems such as plan-
making and improvisation, and the cognitive ning production levels require balancing human
demands of maintaining situational awareness, resources, technologies, inventory, and sales
which place significant demands on attention, (Jonassen, Privish, Christy, & Stavrulaki, 1999).
pattern recognition, and working memory Classical situated case problems also exist in
(Durso & Gronlund, 1999). Skills that are impor- international relations, such as "given low crop
tant to air traffic controllers, for instance, include productivity in the Soviet Union, how would the
the ability to prioritize, to plan, to execute, to solver go about improving crop productivity if
think ahead, to concentrate, and to deal with he or she served as Director of the Ministry of
dynamic visual movements, as well as good sit- Agriculture in the Soviet Union" (Voss and Post,
uational awareness, short term memory, deci- 1988, p. 273). International relations problems
siveness, and perceptual speed and accuracy involve decision making and solution genera-
(Heil, 1999). The multifaceted nature of strategic tion and testing in a political context.
performances makes them especially difficult, In these ill-structured problems, goals are
however, these cognitive demands are situation- vaguely defined; no constraints may be stated,
ally specific. Arguing a case in court, for little is known about how to solve the problem;
instance, would demand a different set of cogni- there is no consensual agreement on what con-
tive skills from those needed for air traffic con- stitutes a good solution; and information avail-
trolling. able to the problem solver is prodigious but
Typically, a finite number of tactical activities incomplete, inaccurate, or ambiguous (Voss et
have been designed to accomplish the strategy, al., 1991). Therefore, "the whole process of cop-
however, the mark of an expert strategic per- ing with a complex problem can be seen as a
former is the ability to improvise or construct process of intention regulation" (D6rner &
new tactics on the spot to meet the strategy. The Wearing, 1995). To complicate the process,
quarterback who calls an audible at the line of "there are no formal procedures or guidelines to
scrimmage is selecting a new tactic to meet the govern case analysis or evaluation of problem
offensive strategy. In battlefield situations, supe- solutions," and what skilled performers need to
rior officers identify a strategy and may negoti- know in order to solve these complex case prob-
ate tactical concerns with the performer, lems is often tacit (Wagner, 1991, p. 179).
however, both realize that tactics may have to be Case-analysis problems generally engage a
~0 ETR&D,Vol. 48, No. 4

process that includes goal elaborating, informa- istics of design problems, including many
tion collecting, hypothesis forming, forecasting degrees of freedom in the problem statement,
(predicting effects), planning and decision mak- which consists only of goals and intentions, lim-
ing, monitoring the effects of ones' actions, and ited or delayed feedback from the world, arti-
self-reflecting (DSrner & Wearing, 1995). That facts as outputs that must function
process is likely to change depending on the independently of the designer, and answers that
nature and context of a problem. Case-analysis tend to be neither right nor wrong, only better or
problems are the most contextually bound kind worse. The importance of an artifact as evidence
of problem, so analyzing cases places a great of problem solving and the lack of dear stan-
importance on situation analysis. dards for evaluating solutions are what make
design problems so ill-structured. Because of the
very ill-structured and complex nature of design
problems, they require the problem solver to
Design Problems
engage in extensive problem structuring, often
using artificial symbol systems (Goel & Pirolli,
Design problems are usually among the most 1989). They also require greater commitment
complex and ill-structured kinds of problems and self-regulation by the problem solver.
that are encountered in practice. For many
Designing greatly exceeds the normal con-
years, researchers (Reitman, 1965; Simon, 1973)
cept of transfer. Designers must structure the
have characterized design problems as ill-struc-
problem by defining the nature of the artifact
tured because they have ambiguous specifica-
that will satisfy the ill-defined requirements.
tion of goals, no determined solution path, and
Because the criteria for acceptable solutions are
the need to integrate multiple knowledge
not always obvious, designers must construct
domains. Whether it be an electronic circuit, a
personalized systems for evaluating their prod-
house, a new entr6e for a restaurant, a musical
ucts. Designers are aided in the design process
composition, an essay, or any other product or
by experience-based design schemas that
system, designing requires applying general
include components that partition the problem
and domain-specific schemas as well as proce-
into a set of meaningful tasks; components that
dural knowledge. Expert writers, for instance,
assure that tasks will function properly; pro-
are problem solvers who exert tremendous
cesses that control the generation of designs; and
"mental effort in the elaboration, the coordina-
evaluation procedures that ensure effective utili-
tion, and the execution of complex goals and
zation of knowledge (Jeffries et al., 1981). As with
subgoals, such as how to shape content for a par-
most problems, these problem schemas are used
ticular audience, how to express conceptual
to monitor and regulate performance. Despite
intentions in the language of prose, or how to
their ill-structuredness, design problems may be
construct a catchy title" (Bryson, Bereiter,
the most important type of problem to investi-
Scardamalia, & Joram , 1991, p. 61). Unfortu-
gate because so many professionals get paid for
nately, most of the research on design problems
designing things (products, systems, etc.).
has been in the area of software design, which is
more well-structured than most design prob-
lems. Early research attempted to apply tradi-
tional top-down approaches to task Dilemmas
decomposition (Jeffries, Turner, Poison, &
Atwood, 1981), however the design process Everyday life is replete with personal, social
deviates significantly from a top-down and ethical dilemmas. Although appearing often
approach (Guindon, 1990). Because software as decision-making problems (Should abortions
design is constrained by language and systems, be banned?), dilemmas are the most ill-struc-
and the task is less ambiguous, it is not a gener- tured and unpredictable type because often
alizable model for patterning design problems. there is no solution that is satisfying or accept-
Why are design problems so ill-structured? able to most people, and there are compromises
Goel and Pirolli (1989) articulated the character- implicit in every solution.
TOWARD A DESIGN THEORY OF PROBLEMSOLVING 81

Two of the most commonly researched d e s c r i e d learning enterprises. For purposes of

dilemmas are social dilemmas and ethical analysis, the typology of problems that I just
dilemmas. Social dilemmas, like resource man- described assumes that these problems are dis-
agement and pollution dilemmas, are perceived crete, that is, each problem of any type com-
by individuals in terms of their own personal prises a single learning objective and is learned
self-interests or in terms of the common good in isolation from another. However, that does
(Schroeder, 1995). Most people select solutions not imply that problems occur as discrete skills.
to dilemmas that serve their self-interests, but On the contrary, problems that are encountered
few seek solutions that serve the common good. in everyday and professional contexts often con-
The larger the group that is involved in any sist of combinations of problems. Activity sys-
dilemma, the more likely that individuals will tems in everyday and professional contexts
pursue their self-interests, because the cost to engage people in solving complex combinations
any one person of providing for the common of well-structured and ill-structured problems:
good exceeds the benefits that the individual what I refer to as metaproblems. Metaproblems
would receive (Olson, 1965). consist not only of clusters of interrelated prob-
Dilemmas are often more complex than the lems related to the same work activities but also
dichotomous social ones just described. For of the regulatory and reflective skills to monitor
example, ethical dilemmas are often compli- and solve combinations of problems. For exam-
cated when the decision maker is pulled in one ple, developing a computer system requires
direction by ethical considerations and in solving a host of design, troubleshooting, and
another direction by legal temporal, or organi- case problems. Running a retail business like-
zational obstacles (MacKay & O'Neill, 1992). The wise represents a myriad of decision-making
current crisis In Kosovo is a prime example of a problems. Problems in everyday and profes-
large-scale dilemma that involves a plethora of sional contexts are generally metaproblems, so
interacting historical political military, reli- when analyzing any problem context, it is neces-
gious, ethical, economic, and anthropological sary to identify not only the problems that are
issues. While each individual in any dilemma solved but also how they are clustered into
may see the problem as dichotomous, there are metaproblems. The metacognitive skills
so many perspectives on the situation that none required to regulate the solution of
is able to offer a generally acceptable solution to metaproblems need to be identified and elabo-
the crisis. The situation is so complex and unpre- rated.
dictable that no best solution can ever be known.
That does not mean that there are not many
solutions, which can be attempted with varying CONCLUSION
degrees of success, however, none will ever
meet the needs of the majority of people or
In this article, I have defined problems and artic-
escape the prospects of catastrophe. Dilemmas
ulated variations in the nature of problems and
are often complex social situations with conflict-
problem representations. I have also briefly
ing perspectives, and they are usually the most
described individual differences that affect the
vexing of problems.
nature of problem-solving activities. Finally, I
have described a typology (and to a lesser
degree, a taxonomy) of problems, including log-
Discrete Problems vs. Metaproblems ical problems, algorithmic problems, story prob-
lems, rule-using problems, decision-making
Most instructional design models provide problems, troubleshooting problems, diagnosis-
microlevel prescriptions, that is, prescriptions solution problems, strategic performance, case-
for how to learn only a single objective or goal analysis problems, design problems, dilemmas,
(Reigeluth, 1983). In an effort to accommodate and metaproblems.
the demands of more complex combinations of If we believe that the cognitive and affective
tasks or objectives, Gagn~ and Merrill (1990) requirements of solving different kinds of prob-
82 ETR&D,Vol. 48, No. 4

lems vary, then so, too, must the nature of sented in this paper is not promulgated as a
instruction that we use to support the develop- definitive theory, but rather as a work in prog-
ment of problem-solving skills. Why? Among ress. Experimentation, assessment, and diaiogue
the most fundamental beliefs of instructional about these problem types and the forthcoming
design is that different learning outcomes models are needed to validate anything
require different instructional conditions approaching a definitive theory for problem-
(Gagn~, 1980). So, the rationale for attempting to solving instruction. V1
articulate different kinds of problem solving in
this paper is to begin to prescribe instructional
analysis and design processes. David H. Jonassen is Distinguished Professor of
Information Scienceand Learning Technologiesat
From the research, it appears obvious that the the University of Missouri. Comments may be
key to learning to solve problems is the problem addressed to [email protected]~
space construction, because rich problem repre-
sentations most dearly distinguish experts from
novices and scaffold working memory (an REFERENCES
essential cognitive component in problem solv-
Anderson, J.R. (1980).Cognitive psychology and its impli-
ing). So, developing elaborate, multiple repre- cations. New York: Freeman.
sentations of problems along with learning to Barrows, H.S. (1985).How to design a problem-based cur-
regulate different kinds of problem performance riculum for the pre-clinical years. New York:Springer.
needs to be explicitly taught. Articulating a set Barrows, H.S., & Tamblyn, ILM. (1980). Problem-based
of strategies to support problem space construc- learning: An approach to medical education. New York:
Springer.
tion is the focus of future research. Bransford, J., & Stein, B.S. (1984). The IDEAL problem
solver: A guidefor improving thinking, learning, and cre-
ativity. New York:W.H. Freeman. Bryson, M., Berei-
FUTURE DEVELOPMENTS ter, C., Scardamalia, M., & Joram, E. (1991). Going
beyond the problem as given: Problem solving in
Currently, research is ongoing to validate this expert and novice writers. In R.J. Sternberg & P.A.
problem typology and articulate cognitive, Frensch (Eds.), Complex problem solving: Principles
and mechanisms (pp. 61-84). Hillsdale, NJ: Lawrence
affective, and conative processes required to
Erlbaum Associates.
solve each type of problem. Those processes will Davidson, J.E., & Sternberg, R.J. (1998). Smart problem
constitute the content of the next paper in this solving:How metacognition helps. In D.J. Hacker, J.
series. If instructional technologists are able to Duniosky, & A.C. Graesser (Eds.), Metacognition in
articulate each problem type, its representation, educational theory and practice (pp. 47--68).Mahwah,
and the kinds of cognitive and affective pro- NJ: LawrenceErlbaum Associates.
Davis, J.K., & Haueisen, W.C. (1976). Field indepen-
cesses that are required to solve each kind of dence and hypothesis testing. Perceptual and Motor
problem, then we should be able also to project Skills, 43, 763-769.
the kinds of individual differences that are likely D6mer, D., & Wearing, A.J. (1995). Complex problem
to interact with those processes, as well as mod- solving:Toward a theory. In P.A. Frensch &J. Funke
els and methods of instructional design, to sup- (Eds.), Complex problem solving: The European perspec-
tive (pp. 65-99). Hillsdale, NJ: Lawrence Erlbaum
port learning of each type of problem. Those Associates.
models will constitute the third paper in the Dunkle, M.E., Schraw, G., & Bendixen, L.D. (1995,
series. Although many instructional methods April). Cognitive processes in well-defined and ill-
between models are likely to overlap, each class defined problem solving. Paper presented at the annual
should suggest a different combination and meeting of the American Educational Research
Association,San Francisco,CA.
sequence of instructional transactions. These dif-
Durso, F.T., & Gronlund, S.D. (1999). Situation aware-
ferent instructional models can be used to ness. In F.T. Durso (Ed.), Handbook of applied cogni-
design problem-solving instruction and to tion (pp. 283-314). Chichester, UK: John Wiley.
assess the effectiveness of different instructional English, L.D. (1998). Children's reasoning in solving
components for learning to solve problems and relational problems of deduction. Thinking & Reason-
ing, 4(3), 249-281.
to transfer those skills to new problems.
Fishbein, D.D., Eckart, T., Lauver, E., van Leeuwen, R.,
It is important to note that the typology pre- & Langemeyer, D. (1990). Learners" questions and
TOWARD A DESIGNTHEORYOF PROBLEMSOLVING 83

comprehension in a tutoring system. Journal of Edu- Hayes, J.R., & Simon, H.A. (1977). Psychological differ-
cational Psychology, 82,163-170. ences among problem isomorphs. In N.J. Castellan,
Flavell, J.H. (1979). Metacognition and comprehension D.B. Pisoni, & G.R. Ports (Eds.), Cognitive theory (pp.
monitoring: A new era of cognitive development 21-41). Hillsdale, NJ: Lawrence Erlbaum Associates.
i n q ~ . American Psychologist, 34, 906-911. Heil, M.C. (1999). Air traffic control specialist age and cog-
Forgas, J.P. (1982). Reactions to life dilemmas: Risk tak- nitive test performance. FAA Office of Aviation Medi-
ing, success and responsibility attribution. Austra- cine Report No. DOT-FAA-AM-99-23. Oklahoma
lian Journal of Psychology, 34, 25-35. City, OK: Federal Aviation Administration.
Funke, J. (1991). Solving complex problems: Explora- Heller, L.C. (1982). An exploration of the effect of
tion and control of complex systems. In RJ. Stem- structure variables on mathematical word problem-
berg & P.A. Frensch (eds.), Complex problem solving: solving achievement (Doctoral dissertation, Rutgers
Principles and mechanisms (pp. 185-222). Hiilsdale, University), Dissertation Abstracts International, 44,
NJ: Lawrence Erlbaum Associates. 416.
Gagn~, R.M. (1980). The conditions of learning. New Hofer, B.K., & Pintrich, P.R. (1997). The development
York: Holt, Rinehart, & Winston. of epistemologicaltheories: Beliefs about knowledge
Gagn~, R.M., Briggs, L.J., & Wager, W.W. (1992). Prin- and knowing and their relation to learning. Review of
ciples of instructional design (4th Ed.). New York: Har- Educational Research, 67(1), 88-140.
court, Brace, & Jovanovich. Hong, N.S., Jonassen, D.H., & McGee, S. (in press).
Gagn~, R.M., & Merrill, M.D. (1990). Integrative goals Predictors of well-structured and ill-structured
for instructional design. Educational Technology problem solving in an astronomy simulation. Journal
Research & Development, 38(1), 23-30 1990. of Research in Science Teaching.
Gick, M.L. (1986). Problem-solving strategies. Educa- Jeffries, R., Turner, A.A., Polson, P.G., & Atwood, M.E.
tional Psychologist, 21 (1&2), 99-120. (1981). The processes involved in designing soft-
Gick, M.L., & Holyoak, K.J. (1980). Analogicalproblem ware. In J.R. Anderson (Ed.), Cognitive skills and their
solving. Cognitive Psychology, 12, 306-355. acquisition (pp. 255-283). Hillsdale, NJ: Lawrence
Gick, M,L., & Holyoak, K.J. (1983). Schema induction Erlbaum Associates.
and analogical transfer. Cognitive Psychology, 15, 1- Johnson, S.D. (1988). Cognitive analysis of expert and
38. novice troubleshooting performance. Performance
Goel, V., & Pirolli, P. (1989). Motivating the notion of Improvement Quarterly, 1(3), 38-54.
generic design within information processing the- Johnson, S.D., & Satchwell, S.E. (1993). The effect of
ory: The design problem space. AI Magazine, 10(1), functional flow diagrams on apprentice aircraft
19-36. mechanics' technical system understanding. Perfor-
Gordon, S.E., & Gill, R.T. (1989). Theformation and use of mance Improvement Quarterly, 6(4), 73-91.
knowledge structures in problem solving domains. Tech. Jonassen, D.H. (1997). Instructional design model for
Report AFOSR-88-0063. Washington, DC: Boiling well-structured and ill-structured problem-solving
AFB. learning outcomes. Educational Technology Research
Gourgey, A.F. (1998). Metacognition and basic skills and Development 45(1), 65-95.
instruction. Instructional Science, 26, 81-96. Jonassen, D.H. (2000a). Integrating problem solving
Greeno, J. (1978). Natures of problem-solving abilities. into instructional design. In R.A. Reiser & J.
In W. Estes (Ed.), Handbook of learning and cognitive Dempsey (Eds.), Trends and issues in instructional
processes (pp. 239-270). Hillsdale, NJ: Lawrence design and technology. Upper Saddle River, NJ: Pren-
Erlbaum Associates. tice-Hall.
Greeno, J. (1991). A view of mathematical problem Jonassen, D.H. (2000b). Activity theory revisited. In
solving in school. In M.U. Smith (Ed.), Toward a uni- D.H Jonassen & S.L. Land, (Eds.), Theoreticalfounda-
fied theory of problem solving (pp. 69-98). Hillsdale, tions of learning environments. Mahwah, NJ: Law-
NJ: Lawrence Erlbaum Associates. rence Erlbaum Associates.
Guindon, R. (1990). Designing the design process: Jonassen, D.H. (2000c). Using technologies to model stu-
Exploiting opportunistic thoughts. Human-Computer dent problem spaces. Paper presented at the Interna-
Interaction, 5, 305-344. tional Conference on Computers in Education,
Halgren, S.L., & Cooke, N.J. (1993). Towards ecological Taipei, Taiwan.
validity in menu research. International Journal of Jonassen, D.H., Beissner, K., & Yacci, M. (1993). Struc-
Man-Machine Studies, 39(1), 51-70. tural knowledge: Techniques for assessing, conveying,
Hail, E.P., Gott, S.P., & Pokomy, R.A. (1995). A proce- and acquiring structural knowledge. HiUsdale, NJ:
dural guide to cognitive task analysis: The PARI method- Lawrence Erlbaum.
o/ogy, Tech. Report AL/HR-TR-1995-0108. Brooks Jonassen, D.H., & Grabowski, B.L. (1993). Handbook of
Air Force Base, TX: Human Resources Directorate. individual d~j:erences, learning and instruction.
Hannafin, MJ., Hall, C., Land, S., & Hill, J. (1994). Hiilsdale, NJ: Lawrence Erlbaum Associates.
Learning in open-ended learning environments: Jonassen, D.H., & Henning, P. (1999). Mental models:
Assumptions, methods, and implications. Educa- Knowledge in the head and knowledge in the world.
tional Technology, 34(8), 48-55. Educational Technology, 39(3), 37-42.
84 ETR&D. Vol. 48, No. 4

Jonassen, D.H., & Kwon, H.I. (in press). Communica- Masui, C., & DeCorte, E. (1999). Enhancing learning
tion patterns in computer-mediated vs. face-to-face and problem solving skills: Orienting and self-judg-
group problem solving. Educational Technology ing, two powerful and trainable learning skills.
Research and Development. Learning and Instruction, 9, 517-542.
Jonassen, D.H, & Land, S.L (Ecls.). (2000). Theoretical Mayer, R.E. (1992). Thinking, problem solving, cognition
foundations of learning environments. Mahwah, NJ: (2nd ed.). New York: Freeman.
Lawrence Erlbaum Associates. Mayer, R.E. (1998). Cognitive. Metacognitive, and
Jonassen, D., Prevish, T., Christy, D., Stavurlaki, E. motivational aspects of problem solving. Instruc-
(1999). Learning to solve problems on the Web: tional Science, 26, 49-63.
Aggregate planning in a business management Mayer, R.E., & Wittrock, M.C. (1996) Problem-solving
course. Distance Education: An International Journal, transfer. In D.C. Berlinert & R.C. Ca]fee (Eds.), Hand-
20(1), 49-63. book of educational psychology (pp. 47-62). New York:
Jonassen, D.H., & Tessmer, M. (1996/1997). An out- Macmillan.
comes-based taxonomy for instructional systems McCloskey, M., Caramaza, A., & Basili, A. (1985). Cog-
design, evaluation, and research. Training Research nitive mechanisms in number processing and calcu-
Journal, 2,11-46. lation: Evidence from dyscalculia. Brain and
Jonassen, D.H., Tessmer, M., & Hannum, W. (1999). Cognition, 4,171-196.
Handbook of task analysis procedures. Mahwah, NJ: Meacham, J.A., Emont, N.C. (1989). The interpersonal
Lawrence Erlbaum Associates.. basis of everyday problem solving. In J.D. Sinnott
Kahney, H. (1993). Problem solving: Current issues. (Ed.), Everyday problem solving: Theory and applica-
Buckingham, UK: Open University Press. tions (pp. 7-23). New York: Praeger.
Kerstholt, J.H., & Raaijmakers, J.G.W. (1997). Decision Mullen, J.D., & Roth, B.M. (1991). Decision making: Its
making in dynamic task environments. In R. Ran- logic and practice. Savage, MD: Rowman & Littlefield.
yard, W.R. Cozier, & Ola Swenson (Ecls.), Decision Newel], A., & Simon, H. (1972). Human problem solving.
making: Cognitive models and explanations (pp. 205- Englewood Cliffs, NJ: Prentice Hall.
217). London: Routledge. Olson, M. (1965). The logic of collective action. Cam-
Kitchner, K.S. (1983). Cognition, metacognition, and bridge, MA: Harvard University Press.
epistemic cognition: A three-level model of cogni- Perkins, D.N., Hancock, C., Hobbs, R., Martin, F., &
tive processing. Human Development, 26, 222-232. Simmons, R. (1986). Conditions of learning in novice
Kluwe, R.H. (1995). Single case studies and models of programmers. Journal of Educational Computing
complex problem solving In P.A. Frensch & J. Funke Research, 2(1), 37-56.
(Eds.), Complex problem solving: The European perspec- Perry, W.G. (1970). The forms of intellectual and ethical
tive (pp. 269-291). Hil]sdale, NJ: Lawrence Erlbaum development in the college years: A scheme. San Fran-
Associates. cisco: Jossey-Bass.
Land, S.M., & Hannafin, M.J. (1996). A conceptual Pokorny, R.A., Hal], E.P., Ga]laway, M.A., & Dibble, E.
framework for the development of theories-in- (1996). Analyzing components of work samples to
action with open-ended learning environments. evaluate performance. Military Psychology, 8(3), 161-
Educational Technology Research& Development, 44(3), 177.
37-53. Reed, S.K., Ernsyt, G.W., & Banerji, R. (1974). The role
Lehman, D., Lempert, R., & Nisbett, R.E. (1988). The of analogy in transfer between similar problem
effects of graduate training on reasoning: Formal states. Cognitive Psychology, 6, 436-450.
discipline and thinking about everyday-life events. Reigeluth, C.M. (1983). Instructional design theories and
Educational Psychologist, 43, 431--42. models. Hillsdale, NJ: Lawrence Erlbaum Associates.
Lester, F.K. (1994). Musings about mathematical prob- Reitman, W.R. (1965). Cognition and thought. New
lem-solving research: 1970-1994. Journalfor Research York: Wiley.
in Mathematics Education, 25, 660-675. Robertson, W.C. (1990). Detection of cognitive struc-
Lucangelli, D., Tressoldi, P.E., & Cendron, M. (1998). ture with protocol data: Predicting performance on
Cognitive and metacognitive abilities involved in physics transfer problems. Cognitive Science, 14, 253-
the solution of mathematical word problems: Vali- 280.
dation of a comprehensive model. Contemporary Ronning, McCurdy, & Ballinger (1984, January). Indi-
Educational Psychology, 23, 257-275. vidual differences: A third component in problem-
MacKay, E., & O'Neill, P. (1992). What creaes the solving instruction. Journal of Research in Science
dilemma in ethical dilemmas? Examples from psy- Teaching, 21(1), 71-82.
chological practice. Ethics & Behavior, 2(4), 227-244. Salomon, G., & Perkins, D.N. (1989). Rocky roads to
Maloney, T.J. (1981). The relation between field-inde- transfer: Rethinking mechanisms of a neglected phe-
pendence and rule-transfer (Doctoral dissertation, nomenon. Educational Psychologist, 24,113-142.
University of Toledo), Dissertation Abstracts Interna- Schacter, J., Chung, G.K.W.K., & Dorr, A. (1998).
tional, 442, 2575. Children's Internet searching on complex problems:
Marshall, S.P. (1995). Schemas in problem solving. Cam- Performance and process analyses. Journal of the
bridge, UK: Cambridge University Press. American Societyfor Information Science, 49, 840-849.
TOWARD A DESIGNTHEORYOF PROBLEMSOLVING 85

Schank, R.C., Fano, A., Bell, B., & Jona, M. (1993/1994). 177-190.
The design of goal-based scenarios. The Journal of the Sweller, J. (1988). Cognitive load during problem solv-
Learning Sciences, 3(4), 305-345. ing: Effects on learning. Cognitive Science, 12, 257-
Schroeder, D.A. (1995). An introduction to social 285.
dilemmas. In D.A. Schroeder (Ed.), Social dilemmas: Tversky, A., & Kahneman, D. (1981). The framing of
Perspectives on individuals and groups (pp. 1-14). decisions and the psychology of choice. Science, 211,
Westport, CT: Praeger. 453-458.
Shavelson, ILJ. (1972). Some aspects of the correspond-
ence between content structure and cognitive struc- van Merri'enboer, J.J.G. (1997). Training complex cogni-
ture in physics instruction. Journal of Educational tive skills. Englewood Cliffs, NJ: Educational Tech-
Psychology, 63, 22,5-234. nology Publications.
Sherrill, J.M. (1983). Solving textbook mathematical Voss, J.F., & Post, T.A. (1988). On the solving of ill-
problems. Alberta Journal of Educational Research, 29, structured problems. In M.T.H. Chi, R. Glaser, &
140-152. M.J. Fan- (Eds.), The nature afexpertise. Hillsdale, NJ:
Simon, H.A. (1973). The structure of ill-structured Lawrence Erlbaum Associates.
problems. Art~qcial Intelligence, 4,181-201. Voss, J.F., Wolfe, C.R., Lawrence, J.A, & Engle, J.A
Simon, D.P. (1978). Information processing theory of (1991). From representation to decision: An analysis
human problem solving. In D. Estes (Ed.), Handbook of problem solving in international relations. In R.J.
of learning and cognitive process. Hillsdale, NJ: Law- Sternberg & P.A. Frensch (Eds.), Complex problem
rence Erlbaum Associates. solving: Principles and mechanisms (pp. 119-158).
Singley, M.K., & Anderson, J.R. (1989). The transfer of Hillsdale, NJ: Lawrence Erlbaum Associates.
cognitive skill. Cambridge, MA: Harvard University Wagner, R.K. (1991). Managerial problem solving. In
Press. R.J. Sternberg & P.A. Frensch (Eds.), Complex problem
Smith, M.U. (1991). A view from biology. In M.U. solving: Principles and mechanisms (pp. 159-184).
Smith (ed.), Toward a untied theory of problem solving. Hillsdale, NJ: Lawrence Erlbaum Associates.
Hi]lsdale, NJ: Lawrence Erlbaum Associates.
Williams, S. (1992). Putting case-based instruction into
Smith, P.L, & Ragan, T.J. (1999). Instructional design
2nd ed. Columbus, OH: Merrill. context: Examples for legal and medical education~
The Journal of the Learning Sciences, 2(4), 367-427.
Sternberg, R.J., & Prensch, P.A. (Eds.).(1991). Complex
problem solving: Principles and mechanisms. Hilisdale, Wood, P.K. (1983). Inquiring systems and problem
NJ: Lawrence Erlbaum Associates. structures: Implications for cognitive development.
Stewin, L., & Anderson, C. (1974). Cognitive complex- Human Development, 26, 249-265.
ity as a determinant of information processing. Woods, D.R., Hrymak, A.N., Marshall, R.R., Wood,
Alberta Journal of Educational Research,20(3), 233-243. P.E., Crowe, Hoffman, T.W., Wright, J.D., Taylor,
Suedfeld, P., de Vries, B., Bluck, S., Wallbaum, B.C. P.A., Woodhouse, K.A., & Bouchard, C.G.K. (1997).
(1996). Intuitive perceptions of decision-making Developing problem-solving skills: the McMaster
strategy: Naive assessors' concepts of integrative problem solving program. Journal of Engineering
complexity. International Journal of Psychology, 3•(5), Education, 86(2), 75-92.