Chapter 10: Improving Decision Making and Managing Knowledge 353

INTERACTIVE SESSION: TECHNOLOGY IBM’s Watson: Can Computers Replace Humans?

In February 2011, an IBM computer named Watson the clue, and generate hundreds of solutions. Human
took on the two most-decorated champions of the beings don’t need to take such a formal approach
game show Jeopardy, Ken Jennings and Brad Rutter. to generate the solutions that fit a question best, but
Watson, which was named after IBM’s founder, Watson compensates for this with superior computing
Thomas J. Watson, won handily. power and speed. If a certain algorithm works to solve
Watson’s achievement represents a milestone in the a problem, Watson remembers what type of question it
ability of computers to process and interpret human was and the algorithm it used to get the right answer.
language. The goal of the IBM team, a group of 25 In this way, Watson improves at answering questions
researchers led by Dr. David Ferrucci, was to develop over time. Watson also learns another way—the team
a more effective set of techniques that computers can gave Watson thousands of old Jeopardy questions to
use to process “natural language”—language that process. Watson analyzed both questions and answers
human beings instinctively use, not language specially to determine patterns or similarities between clues,
formatted to be understood by computers. Watson had and using these patterns, it assigns varying degrees of
to go far beyond responding to simple commands, confidence to the answers it gives.
or receiving only specific, pre-defined inputs. Watson was able to correctly answer only a
Jeopardy questions are renowned for their wordplay, small fraction of the questions it was given at first,
hidden meanings, and tricky puns, and can deal but machine learning allowed Watson to continue
with practically any subject matter. For computers, improving, first growing closer to Jeopardy contestant
language processing doesn’t get much harder. level, and then finally reaching the Jeopardy champion
Watson’s performance on the show was far from level of Jennings and Rutter. Similar machine learning
perfect. The computer blundered on a Final Jeopardy techniques are used in computer models that predict
question with the category “U.S. Cities,” where it weather, likes and dislikes from online retailers like
selected Toronto as its response. Sometimes, the Amazon.com, and speech recognition to develop more
answers in a category do not directly correspond with accurate predictions, recommendations, and speech
the title of the category, so Watson didn’t assume, as the processing.
human players did, that the answer had to be a city in IBM believes future applications for Watson are
the United States. Nevertheless, Watson overcame that numerous and wide-ranging in medicine, financial
flub to win convincingly, finishing with over $77,000 to services, or any industry where sifting through large
about $20,000 apiece for its human opponents. amounts of data to answer questions is important. In
The hardware required for Watson to work so September 2011, health insurer WellPoint Inc., with
quickly and accurately was staggeringly powerful. 34.2 million members, enlisted Watson to help it
Watson consists of 10 racks of IBM POWER 750 servers choose among treatment options and medicines. The
running Linux, uses 15 terabytes of RAM and 2,880 WellPoint application will combine data from three
processor cores (equivalent to 6,000 top-end home com- sources: a patient’s chart and electronic records main-
puters), and operates at 80 teraflops. Watson needed this tained by a physician or hospital, the insurance compa-
amount of power to quickly scan its enormous database ny’s history of medicines and treatments, and Watson’s
of information, including information from the Internet. huge library of textbooks and medical journals. Watson
The team downloaded over 10 million documents, should be able to process all of the data and answer
including encyclopedias and Wikipedia, the Internet a question in moments, providing several possible
Movie Database (IMDB), and the entire archive of the diagnoses or treatments, ranked in order of the com-
New York Times. All of the data sat in Watson’s primary puter’s confidence, along with the basis for its answer.
memory, as opposed to a much slower hard drive, so It’s unclear how effective Watson will actually
that Watson could find the data it needed within three be in this field, where the information available in
seconds. The Watson project took 20 IBM engineers medical journals and other sources is highly disor-
three years to build at an $18 million labor cost, and an ganized, often contradictory, and littered with typos
estimated $1 million in equipment. and inconsistent naming conventions. When human
Watson is able to learn from its mistakes as well doctors apply their understanding of disease to our
as its successes. To solve a typical problem, Watson bodies, it is based on knowledge of the literature,
tries many of the thousands of algorithms that the but also based on prior experience and good guesses.
team has programmed it to use. The algorithms evalu- There’s no guarantee that Watson can overcome these
ate the language used in each clue, gather information obstacles, but if it can, it could dramatically increase
about the important people and places mentioned in efficiency and accuracy of medical diagnoses.
354 Part III: Key System Applications for the Digital Age

Watson’s ability to process natural language Let’s ask Watson: Take a stab at predicting your
allows it to perform many jobs requiring factual future impact! Watson is silent on this and other
knowledge and expertise. Jobs that involve answering topics. As it turns out, Watson was not programmed to
questions or conducting transactions on the telephone look into the future, or to have intentions, objectives,
are likely candidates for replacement. While hundreds or feelings about the experience of being Watson.
of thousands of people who perform these jobs could
be unemployed, many businesses could use Watson’s Sources: Anna Wilde Mathews, “Wellpoint’s New Hire. What Is Watson?” Wall
Street Journal, September 12, 2011; John Markoff, “A Fight to Win the Future:
technology to increase efficiency and improve their
Computers vs. Humans,” New York Times, February 14, 2011; John Markoff,
bottom lines. “Computer Wins on ‘Jeopardy!’: Trivial, It’s Not,” New York Times, February
But does Watson really understand language or 16, 2011; “IBM’s Watson Heads From ‘Jeopardy!’ To Columbia University
Medical Center,” CBSNewYork.com, February 17, 2011; Stanley Fish, “What
the answers it’s giving? Skeptics of artificial intelli- Did Watson the Computer Do?” New York Times, February 21, 2011; Charles
gence insist that it doesn’t. The IBM researchers who Babcock, “Watson’s Jeopardy Win a Victory For Mankind,” InformationWeek,
designed the system don’t disagree. But to the IBM February 24, 2011; Greg Lindsay,“Changing The Game: ‘How I Beat Watson
and Came Out a Different Player” and Stephen Baker, “The Programmer’s
team, that’s less important than the fact that Watson can Dilemma: Building a Jeopardy! Champion,” McKinsey Quarterly, February
even answer questions correctly as frequently as it does. 2011.

CASE STUDY QUESTIONS MIS IN ACTION

1. How powerful is Watson? Describe its technology. Visit the IBM Web site and search for information on
Why does it require so much powerful hardware? Watson. Then answer the following questions:
2. How “intelligent” is Watson? What can it do? 1. Which industries or disciplines is IBM targeting
What can’t it do? with Watson?
3. What kinds of problems is Watson able to solve? 2. What improvements is IBM hoping to make in
Watson’s basic functionality?
4. Do you think Watson will be as useful in other
disciplines as IBM hopes? Will it be beneficial to 3. Try to find video of Watson playing Jeopardy.
everyone? Explain your answer. What kinds of questions does it get wrong?

Knowledge management increases the ability of the organization to learn from its environ-
ment and to incorporate knowledge into its business processes and decision making.
Knowledge that is not shared and applied to the problems facing firms and managers
does not add any value to the business. Knowing how to do things effectively and efficiently
in ways that other organizations cannot duplicate is a major source of profit and competitive
advantage. Why? Because the knowledge you generate about your own production processes,
and about your customers, usually stays within your firm and cannot be sold or purchased
on the open market. In this sense, self-generated business knowledge is a strategic resource
and can provide strategic advantage. Businesses will operate less effectively and efficiently
if this unique knowledge is not available for decision making and ongoing operations.
There are two major types of knowledge management systems: enterprise-wide knowledge
management systems and knowledge work systems.

ENTERPRISE-WIDE KNOWLEDGE MANAGEMENT SYSTEMS

Firms must deal with at least three kinds of knowledge. Some knowledge exists within
the firm in the form of structured text documents (reports and presentations). Decision
makers also need knowledge that is semistructured, such as e-mail, voice mail, chat room
exchanges, videos, digital pictures, brochures, or bulletin board postings. In still other cases,
there is no formal or digital information of any kind, and the knowledge resides in the heads
of employees. Much of this knowledge is tacit knowledge and is rarely written down.
Chapter 10: Improving Decision Making and Managing Knowledge 355

Enterprise-wide knowledge management systems deal with all three types of

knowledge. Enterprise-wide knowledge management systems are general-purpose, firmwide
systems that collect, store, distribute, and apply digital content and knowledge. These sys-
tems include capabilities for searching for information, storing both structured and unstruc-
tured data, and locating employee expertise within the firm. They also include supporting
technologies such as portals, search engines, collaboration tools, and learning management
systems.

Enterprise Content Management Systems

Businesses today need to organize and manage both structured and semistructured knowledge
assets. Structured knowledge is explicit knowledge that exists in formal documents, as well
as in formal rules that organizations derive by observing experts and their decision-making
behaviors. But, according to experts, at least 80 percent of an organization’s business content
is semistructured or unstructured—information in folders, messages, memos, proposals,
e-mails, graphics, electronic slide presentations, and even videos created in different formats
and stored in many locations.
Enterprise content management systems help organizations manage both types of
information. They have capabilities for knowledge capture, storage, retrieval, distribution,
and preservation to help firms improve their business processes and decisions. Such systems
include corporate repositories of documents, reports, presentations, and best practices, as
well as capabilities for collecting and organizing semistructured knowledge such as e-mail
(see Figure 10.14). Major enterprise content management systems also enable users to access
external sources of information, such as news feeds and research, and to communicate via
e-mail, chat/instant messaging, discussion groups, and videoconferencing.
A key problem in managing knowledge is the creation of an appropriate classifica-
tion scheme to organize information into meaningful categories. Once the categories for
classifying knowledge have been created, each knowledge object needs to be “tagged,” or
classified, so that it can be easily retrieved. Enterprise content management systems have
capabilities for tagging, interfacing with corporate databases where the documents are
stored, and creating an enterprise portal environment for employees to use when searching
for corporate knowledge. Open Text, EMC Documentum, IBM, and Oracle are leading
vendors of enterprise content management software.
Barrick Gold, the world’s leading gold producer, uses Open Text LiveLink Enterprise
Content Management tools to manage the massive amounts of information required for
building mines. The system organizes and stores both structured and unstructured content,
including computer-aided design (CAD) drawings, contracts, engineering data, and pro-
duction reports. If an operational team needs to refer back to the original document, that

Figure 10.14
An Enterprise
Content
Management
System
An enterprise content
management system
has capabilities for clas-
sifying, organizing, and
managing structured
and semistructured
knowledge and making it
available throughout the
enterprise.
356 Part III: Key System Applications for the Digital Age

document is in a single digital repository as opposed to being scattered over multiple sys-
tems. Barrick’s electronic content management system reduces the amount of time required
to search for documents, shortening project schedules, improving the quality of decisions,
and minimizing rework (Open Text, 2011).
Firms in publishing, advertising, broadcasting, and entertainment have special needs
for storing and managing unstructured digital data such as photographs, graphic images,
video, and audio content. Digital asset management systems help them classify, store, and
distribute these digital objects.

Knowledge Network Systems

Knowledge network systems, also known as expertise location and management systems,
address the problem that arises when the appropriate knowledge is not in the form of a digital
document but instead resides in the memory of expert individuals in the firm. Knowledge
network systems provide an online directory of corporate experts in well-defined knowledge
domains and use communication technologies to make it easy for employees to find the
appropriate expert in a company. Some knowledge network systems go further by systematizing
the solutions developed by experts and then storing the solutions in a knowledge database as a
best-practices or frequently asked questions (FAQs) repository (see Figure 10.15). Hivemine's
AskMe provides stand-alone knowledge networking capabilities, and some knowledge net-
working capabilities can be found in the leading collaboration software suites.

Collaboration Tools and Learning Management Systems

We have already discussed the role of collaboration tools in information sharing and
teamwork in Chapters 2 and 6. Social bookmarking and learning management systems
feature additional capabilities for sharing and managing knowledge.

Figure 10.15
An Enterprise
Knowledge
Network System
A knowledge network
maintains a database
of firm experts, as well
as accepted solutions
to known problems,
and then facilitates the
communication between
employees looking for
knowledge and experts
who have that knowl-
edge. Solutions created
in this communication
are then added to a
database of solutions in
the form of frequently
asked questions (FAQs),
best practices, or other
documents.
Chapter 10: Improving Decision Making and Managing Knowledge 357

Social bookmarking makes it easier to search for and share information by allowing
users to save their bookmarks to Web pages on a public Web site and tag these book-
marks with keywords. These tags can be used to organize and search for the documents.
Lists of tags can be shared with other people to help them find information of interest.
The user-created taxonomies created for shared bookmarks and social tagging are called
folksonomies. Delicious and Digg are two popular social bookmarking sites.
Suppose, for example, that you are on a corporate team researching wind power. If
you did a Web search and found relevant Web pages on wind power, you would click on a
bookmarking button on a social bookmarking site and create a tag identifying each Web
document you found to link it to wind power. By clicking on the “tags” button at the social
networking site, you would be able to see a list of all the tags you created and select the
documents you need.
Companies need ways to keep track of and manage employee learning and to integrate
it more fully into their knowledge management and other corporate systems. A learning
management system (LMS) provides tools for the management, delivery, tracking, and
assessment of various types of employee learning and training.
For example, the Whirlpool Corporation uses CERTPOINT’s learning management
system to manage the registration, scheduling, reporting, and content for its training
programs for 3,500 salespeople. The system helps Whirlpool tailor course content to the
right audience, track the people who took courses and their scores, and compile metrics on
employee performance.

KNOWLEDGE WORK SYSTEMS

The enterprise-wide knowledge systems we have just described provide a wide range of
capabilities used by many, if not all, the workers and groups in an organization. Firms also
have specialized systems for knowledge workers to help them create new knowledge for
improving the firm’s business processes and decision making. Knowledge work systems
(KWS) are specialized systems for engineers, scientists, and other knowledge workers that
are designed to promote the creation of knowledge and to ensure that new knowledge and
technical expertise are properly integrated into the business.

Requirements of Knowledge Work Systems

Knowledge work systems give knowledge workers the specialized tools they need, such
as powerful graphics, analytical tools, and communications and document management.
These systems require great computing power to handle the sophisticated graphics or
complex calculations necessary for such knowledge workers as scientific researchers,
product designers, and financial analysts. Because knowledge workers are so focused
on knowledge in the external world, these systems also must give the worker quick and
easy access to external databases. They typically feature user-friendly interfaces that
enable users to perform needed tasks without having to spend a lot of time learning
how to use the computer. Figure 10.16 summarizes the requirements of knowledge work
systems.
Knowledge workstations often are designed and optimized for the specific tasks to be
performed. Design engineers need graphics with enough power to handle three-dimensional
CAD systems. However, financial analysts are more interested in access to a myriad of
external databases and technology for efficiently storing and accessing massive amounts of
financial data.

Examples of Knowledge Work Systems

Major knowledge work applications include CAD systems (which we introduced in
Chapter 3), virtual reality systems for simulation and modeling, and financial workstations.
Contemporary CAD systems are capable of generating realistic-looking three-dimen-
sional graphic designs that can be rotated and viewed from all sides. Troy Lee Designs,
which makes sports helmets, recently invested in CAD software that could create the
358 Part III: Key System Applications for the Digital Age

Figure 10.16
Requirements of
Knowledge Work
Systems
Knowledge work
systems require strong
links to external knowl-
edge bases in addition
to specialized hardware
and software.

helmets in 3-D. The technology defined the shapes better than traditional methods, which
involved sketching an idea on paper, hand-molding a clay model, and shipping the model
to Asian factories to create a plastic prototype. Production is now about six months faster
and about 35 percent cheaper, with Asian factories able to produce an exact replica after
receiving the digital design via e-mail (Maltby, 2010).
Virtual reality systems use interactive graphics software to create computer-generated
simulations that are so close to reality that users almost believe they are participating in
a real-world situation. In many virtual reality systems, the user dons special clothing,
headgear, and equipment, depending on the application. The clothing contains sensors
that record the user’s movements and immediately transmit that information back to the
computer. For instance, to walk through a virtual reality simulation of a house, you would
need garb that monitors the movement of your feet, hands, and head. You also would need
goggles containing video screens and sometimes audio attachments and feeling gloves so
that you are immersed in the computer feedback.
A virtual reality system helps mechanics in Boeing Company’s 25-day training course
for its 787 Dreamliner learn to fix all kinds of problems, from broken lights in the cabin
to major glitches with flight controls. Using both laptop and desktop computers inside a
classroom with huge wall-mounted diagrams, Boeing airline mechanics train on a system
that displays an interactive Boeing 787 cockpit, as well as a 3-D exterior of the plane. The
mechanics “walk” around the jet by clicking a mouse, open virtual maintenance access
panels, and go inside the plane to repair and replace parts (Sanders, 2010).
Augmented reality (AR) is a related technology for enhancing visualization.
AR provides a live direct or indirect view of a physical real-world environment whose
elements are augmented by virtual computer-generated imagery. The user is grounded in
the real physical world, and the virtual images are merged with the real view to create the
augmented display. The digital technology provides additional information to enhance the
perception of reality, making the surrounding real world of the user more interactive and
meaningful. The yellow first-down markers shown on televised football games are exam-
ples of augmented reality as are medical procedures like image-guided surgery, where
data acquired from computerized tomography (CT) and magnetic resonance imaging
(MRI) scans or from ultrasound imaging are superimposed on the patient in the operating
room. Other industries where AR has caught on include military training, engineering
design, robotics, and consumer design.
Virtual reality applications developed for the Web use a standard called Virtual
Reality Modeling Language (VRML). VRML is a set of specifications for interactive,
Chapter 10: Improving Decision Making and Managing Knowledge 359

three-dimensional modeling on the World Wide Web that organizes multiple media types,
including animation, images, and audio, to put users in a simulated real-world environ-
ment. VRML is platform independent, operates over a desktop computer, and requires
little bandwidth.
DuPont, the Wilmington, Delaware, chemical company, created a VRML application
called HyperPlant, which enables users to access three-dimensional data over the Internet
using Web browser software. Engineers can go through three-dimensional models as if they
were physically walking through a plant, viewing objects at eye level. This level of detail
reduces the number of mistakes they make during construction of oil rigs, oil plants, and
other structures.
The financial industry is using specialized investment workstations to leverage the
knowledge and time of its brokers, traders, and portfolio managers. Firms such as Merrill
Lynch and UBS Financial Services have installed investment workstations that integrate a
wide range of data from both internal and external sources, including contact management
data, real-time and historical market data, and research reports. Previously, financial
professionals had to spend considerable time accessing data from separate systems and
piecing together the information they needed. By providing one-stop information faster
and with fewer errors, the workstations streamline the entire investment process from stock
selection to updating client records.

LEARNING TRACKS
The following Learning Tracks provide content relevant to topics covered in this
chapter:
1. Building and Using Pivot Tables
2. How an Expert System Inference Engine Works
3. Challenges of Implementing and Using Knowledge Management Systems
4. Business Intelligence

Review Summary

1 What are the different types of decisions, and how does the decision-making
process work? Decisions may be structured, semistructured, or unstructured, with
structured decisions clustering at the operational level of the organization and unstructured
decisions at the strategic level. Decision making can be performed by individuals or groups
and includes employees as well as operational, middle, and senior managers. There are four
stages in decision making: intelligence, design, choice, and implementation.

2 How do business intelligence and business analytics support decision making?

Business intelligence and analytics promise to deliver correct, nearly real-time
information to decision makers, and the analytic tools help them quickly understand the
information and take action. A business intelligence environment consists of data from the
business environment, the BI infrastructure, a BA toolset, managerial users and methods,
a BI delivery platform (MIS, DSS, or ESS), and the user interface. There are six analytic
functionalities that BI systems deliver to achieve these ends: pre-defined production
reports, parameterized reports, dashboards and scorecards, ad hoc queries and searches,
the ability to drill down to detailed views of data, and the ability to model scenarios and
create forecasts.
360 Part III: Key System Applications for the Digital Age

Management information systems (MIS) producing prepackaged production reports are

typically used to support operational and middle management, whose decision making is
fairly structured. For making unstructured decisions, analysts and “super users” employ
decision-support systems (DSS) with powerful analytics and modeling tools, including
spreadsheets and pivot tables. Senior executives making unstructured decisions use
dashboards and visual interfaces displaying key performance information affecting the
overall profitability, success, and strategy of the firm. The balanced scorecard and business
performance management are two methodologies used in designing executive support
systems (ESS).

3 How do information systems help people working in a group make decisions more
efficiently? Group decision-support systems (GDSS) help people meeting together in a
group arrive at decisions more efficiently. GDSS feature special conference room facilities
where participants contribute their ideas using networked computers and software tools for
organizing ideas, gathering information, ranking and setting priorities, and documenting
meeting sessions.

4 What are the business benefits of using intelligent techniques in decision making
and knowledge management? Expert systems capture tacit knowledge from a limited
domain of human expertise and express that knowledge in the form of rules. The strategy to
search through the knowledge base is called the inference engine. Case-based reasoning
represents organizational knowledge as a database of cases that can be continually expanded
and refined.

Fuzzy logic is a software technology for expressing knowledge in the form of rules that
use approximate or subjective values. Neural networks consist of hardware and software
that attempt to mimic the thought processes of the human brain. Neural networks are notable
for their ability to learn without programming and to recognize patterns in massive amounts
of data.

Genetic algorithms develop solutions to particular problems using genetically based

processes, such as fitness, crossover, and mutation. Intelligent agents are software programs
with built-in or learned knowledge bases that carry out specific, repetitive, and predictable
tasks for an individual user, business process, or software application.

5 What types of systems are used for enterprise-wide knowledge management and
knowledge work, and how do they provide value for businesses? Enterprise content
management systems feature databases and tools for organizing and storing structured
documents and semistructured knowledge, such as e-mail or rich media. Knowledge network
systems provide directories and tools for locating firm employees with special expertise
who are important sources of tacit knowledge. Often these systems include group collabora-
tion tools, portals to simplify information access, search tools, and tools for classifying
information based on a taxonomy that is appropriate for the organization. Learning
management systems provide tools for the management, delivery, tracking, and assessment
of various types of employee learning and training.

Knowledge work systems (KWS) support the creation of new knowledge and its
integration into the organization. KWS require easy access to an external knowledge
base; powerful computer hardware that can support software with intensive graphics,
analysis, document management, and communications capabilities; and a user-friendly
interface.