Generative Artificial

Intelligence
A Brief Primer for CARL Institutions

DECEMBER 19, 2023

Table of Contents
Overview ....................................................................................................................................... 2
Examples of widely used general tools..................................................................................................... 2

Research........................................................................................................................................ 3
Examples of literature search and management tools ............................................................................ 4

Teaching and Learning .............................................................................................................. 4

Examples of teaching tools ......................................................................................................................... 5
Examples of writing tools for student learning ....................................................................................... 6

Library Operations ...................................................................................................................... 6

Cataloguing, metadata, discovery and aggregation ............................................................................... 6
Reference ....................................................................................................................................................... 7
Preservation .................................................................................................................................................. 7
Workforce development ............................................................................................................................. 7

Publishing ..................................................................................................................................... 8
Examples of tools for editorial management and peer review .............................................................. 8
Personal work / professional competency............................................................................ 9
Examples of task-based tools...................................................................................................................... 9
Policy ........................................................................................................................................... 10
Misinformation ........................................................................................................................................... 10
Nature of inputs ......................................................................................................................................... 10
Copyright .................................................................................................................................................... 11
Privacy ......................................................................................................................................................... 11
Bias ............................................................................................................................................................... 11
Social impacts ............................................................................................................................................. 11
Conclusion ................................................................................................................................... 12
Further reading........................................................................................................................... 12
Policy and regulation................................................................................................................................. 12
How generative AI tools work ................................................................................................................. 13
Discussions of academic library-related applications and implications ............................................ 13
CARL responses to artificial intelligence issues .................................................................................... 14
Examples of AI LibGuides, Primers, and Other Resources in CARL Institutions ............................ 14
University of Victoria: .............................................................................................................................. 14
Université du Québec à Montréal: ........................................................................................................... 14
Carleton University:................................................................................................................................. 14

CANADIAN ASSOCIATION OF RESEARCH LIBRARIES  BIBLIOTHÈQUES DE RECHERCHE DU CANADA 1

Overview
Artificial intelligence is transforming how we access, interpret, and make use of
information. Recent developments in generative AI, notably the advent of Chat-GPT in
2022, have inspired new applications and consideration within the library realm. The
potential uses of generative AI cross all areas of research library mission and
operation, presenting both opportunities and risks – and currently occasioning much
online analysis, opinion, and hype.

Traditional forms of AI are already in use in many task-specific library applications as

well as in everyday life; the affordances of generative AI have distinct implications.
Traditional AI and generative AI differ in their respective capabilities and applications.
Whereas traditional AI systems are primarily applied to analyze data and make
predictions, generative AI uses its training data to create new content. Though
generative AI is a powerful tool that will only grow in importance, both forms of AI,
with their different capabilities and applications, can work in tandem and are not
mutually exclusive.

This brief document provides a function-based structure to frame consideration of the

various areas of library work that AI will impact. This is a selective snapshot at the
moment, however, and the listings within the outlined areas will soon date. New
applications of AI are emerging all the time.

Library leaders need a critical understanding of generative AI technologies and their

potential uses within library functions, and must encourage all parts of the library
workforce to consider critical AI knowledge as an emerging core competency that
requires fostering, training, and support.

With more knowledge and informed opinion about the technologies, tools, risks and
opportunities, library leaders will be able to contribute usefully to campus-wide
discussions on AI applications and attendant issues.

Examples of widely used general tools

• GPT-4; Claude.ai; Perplexity.ai; Google Bard; Bing Chat
o There are more products than the most-known ChatGPT that serve
similar generative functions. Some reviewers have concluded that others
in this list perform better, in fact. Most offer both free and premium
versions.
• DALL-E 3; Midjourney, Stable Diffusion
o Image generation tools

CANADIAN ASSOCIATION OF RESEARCH LIBRARIES  BIBLIOTHÈQUES DE RECHERCHE DU CANADA 2

 • CORE-GPT (coming soon)
o A GPT-based tool trained on over 32 million full-text open access
scientific articles. It provides evidence-based answers to questions, along
with citations.

Research
Though numerous, the various discipline-specific AI tools that researchers make use of
during the conduct of their research are beyond the scope of this overview. For
example, data gathering, analysis, and management tools are not identified here.

There are existing and emerging opportunities for AI tools to save time for
researchers. Tools like ChatGPT can serve as virtual research assistants, conducting
literature reviews and online experiments, analyzing data, writing grant applications,
abstracts, and papers, generating software code, generating citations, and more.
Libraries can harness the opportunities provided by these tools to better support
researchers and make use of AI-generated analytics to measure impact.

Currently, research technology companies and commercial publishing vendors are

putting significant resources into developing AI-based research tools that serve
researcher workflows across the research and dissemination lifecycle. For example,
Scopus and Web of Science are introducing AI-based conversational interfaces to
their database search functions. These products are typically not transparent and may
be prone to bias. Non-profit organizations are working on open-source tools but are
not as well funded as for-profit vendors. For example, the extent to which OpenAI has
access to publications behind paywalls is unclear.

The added nuances and concerns around AI mean that libraries may want to further
cultivate relationships with vendors, whether through their professional associations
or directly. Libraries need to perform due diligence around a given vendor’s claims of
AI integration, as the capabilities range considerably. As with other library systems,
end users may not be able to tailor the system to local needs to the extent they would
like, and system development and maintenance costs are high. Libraries may need to
make ethics-driven decisions when negotiating with vendors and should be aware of a
given system’s sources of factual data, training data, privacy settings, and other
guardrails.

CANADIAN ASSOCIATION OF RESEARCH LIBRARIES  BIBLIOTHÈQUES DE RECHERCHE DU CANADA 3

Examples of literature search and management tools
• Scite
o A smart citation index that displays the context of citations and classifies
their intent using deep learning. Scite.ai has access to some full text and
publications behind paywalls.
• Consensus
o A search engine that uses AI to extract, aggregate, and distill findings
from research articles.
• Elicit
o An AI tool that finds articles and summarizes article abstracts, while also
mining for keywords/subject headings. Elicit can find relevant papers
without perfect keyword match, summarize takeaways from the papers
specific to your question, and extract key information from the papers.
While Elicit does have access to some full-text and articles behind
paywalls, it may draw direct answers from abstracts alone.
• ResearchRabbit
o A citation-based literature mapping tool, showing links and relationships
between authors and sources. Allows you create collections, categorize
resources, and share collections.
• ChatPDF and Docalysis
o Upload a PDF document and then ask questions about the content of the
PDF.
• Connected Papers
o Provides a visual overview of similar papers to identify trends, popular
works, and the dynamics of a given field.
o Explainpaper
o Synthesizes uploaded passages; designed to provide clear explanations
of dense academic writing.
• Scholarcy
o Summarizes articles; useful features include directing user to Open
Access version on citation and creating a personal collection of
summaries.

Teaching and Learning

AI literacy will be increasingly important as students develop critical thinking skills and
navigate which uses of AI are appropriate for academic work, from brainstorming
through to literature searching, analyzing sources, and writing papers. The ethical
implications related to academic integrity, privacy and assigning/deciphering

CANADIAN ASSOCIATION OF RESEARCH LIBRARIES  BIBLIOTHÈQUES DE RECHERCHE DU CANADA 4

ownership and authorship in student work necessitate developing very clear
guidelines for students.

The competencies to critically evaluate AI technologies and use them reflectively and
ethically complement existing programming related to information, digital and
copyright literacy. Algorithmic literacy – the understanding of one’s interaction with
AI, how AI processes information online, and the use of personal data – is increasingly
vital to all, and libraries are well-positioned to provide instruction. Libraries should be
prepared to answer relevant questions from students and faculty, but they may also
consider proactively promoting their AI expertise, especially to faculty. Libraries must
grow their understanding of prompt engineering and problem formulation so as to
most effectively engage with large language models (LLMs) and provide adequate
instruction to students and faculty in this area.

Some of the opportunities available to libraries are to use AI to support individualized

learning, to develop digital literacy skills, and to support course administrative tasks
such as lesson planning, evaluation and feedback, and the analysis of student success.
But recognized risks include depersonalized learning, the introduction of system-
based biases derived from the AI’s training corpus and decreased critical and creative
thinking skills.

One opportunity for libraries may be in OER production. Tools like ChatGPT can save
writing time for authors, but they will need to (carefully) review and revise the
generated output. The declaration of AI assistance is advisable if human authors are
being credited. Possibly, the reduction in time dedicated to such projects will make
them more attractive for faculty, leading to an increase in production of OER.

Examples of teaching tools

• ClassPoint AI
o Creates quiz questions based on PowerPoint slides
• PowerPoint Speaker Coach
o Helps rehearse presentations, providing feedback based on pacing, pitch,
use of filler words, informal speech, euphemisms, culturally sensitive
terms, etc.
• SlidesAI.io
o Creates attractive and engaging presentation slides
• Synthesia
o AI video generation platform that turns text into videos (can do 120+
languages

CANADIAN ASSOCIATION OF RESEARCH LIBRARIES  BIBLIOTHÈQUES DE RECHERCHE DU CANADA 5

Examples of writing tools for student learning
• Grammarly
o Writing assistance
• Moonbeam
• Helps to organize notes and ideas into well-written content
• QuillBot
o Helps to correct grammar and paraphrase sentences

Library Operations
AI offers the potential to maximize efficiencies in various areas of traditional library
operations. AI tools provide opportunities to extend and enhance existing user
services and their accessibility, and to provide more customized support. In addition
to user services, AI-generated library analytics can support management and
planning.

The selection of appropriate tools, vendors, and partners should be informed by best
practices as they evolve and related codes of conduct, like the Government of Canada
code noted in the Policy section. Libraries may want to consider developing
guidelines with respect to adopting and implementing AI within operational units.

Cataloguing, metadata, discovery and aggregation

In this rapidly evolving environment, the imperative for libraries to optimize their
metadata for machine actionability and interoperability is strengthened.

Cataloguing systems have made use of automation for many years, so it is no surprise
that they would incorporate AI-based tools to accelerate and streamline activity
around cataloguing and metadata. These tools can be especially helpful in handling
the high volume and range of digital resources libraries are now managing; indexing,
generating keywords, document matching and flagging offensive content are just
some examples where AI has utility. Image-based systems employ facial recognition
to identify and create metadata around images.

Library vendors, such as Ex Libris, have declared they are integrating AI tools into
their ILS and other products, such as conversational discovery and article summation,
as part of what they are calling the Next Discovery Experience. OCLC has added AI-
generated book recommendations to WorldCat.org. These new tools have the
potential to streamline workflows and improve access to library collections, as
searchers retrieve more personalized patron results.

CANADIAN ASSOCIATION OF RESEARCH LIBRARIES  BIBLIOTHÈQUES DE RECHERCHE DU CANADA 6

Despite these efficiencies, integrating AI technologies with legacy systems can be
complex and resource-intensive. Libraries should consider data sovereignty issues
when choosing third party providers where processing occurs beyond the local
environment.

Reference
AI tools already answer basic reference questions, freeing up staff time for other
tasks. Current findings suggest that ChatGPT may not yet be adequately trained or
have access to sufficiently current information to perform adequately for a large
research library. However, as training data grows, and vendors develop their own
customized reference chat-bots, AI-generated reference will increase in sophistication
and functionality. The potential exists to offer tailored reference responses based on
user data the library possesses; however, issues around privacy and retention of
personal data mean this area needs to be navigated carefully.

For more complex requests, library staff can help students and researchers use
research tools (such as those listed above) to dig deeper into the literature. Librarians
can leverage their existing research and literacy skills to train students and
researchers in prompt engineering. AI has the capacity to make the investigative
stage of research more efficient and advanced reference is likely to be increasingly
well-served. The expertise and authority of library staff, however, is unlikely to be
replicated by these tools in the near future.

Preservation
AI tools can accelerate and streamline the digitization process, help to identify and
restore damaged or degraded digital artifacts, and improve accessibility. However,
rapid obsolescence and future incompatibility of AI tools for digital preservation
portend future complications.

The advent of “synthetic text” created through AI invites questions for libraries in
terms of what to collect and preserve. The presence of inauthentic content can both
undermine trust in the library and obscure the authoritative historical record.

Workforce development
Libraries should address any “skills gap” amongst library staff to ensure a smooth
integration of AI technologies. Desirable competencies to be maintained and
upgraded include: competency in a wide variety of technologies; understanding of
data science; facility with big data and complex searching; and information, media,
and algorithmic literacies. While AI offers opportunities for library staff to reform and
accelerate aspects of their work, they may feel overwhelmed by expectations around
new competencies.

CANADIAN ASSOCIATION OF RESEARCH LIBRARIES  BIBLIOTHÈQUES DE RECHERCHE DU CANADA 7

Examples of tools serving aspects of library operations
• Microsoft Syntex
o Though not specific to libraries, this tool may be useful to cataloguers for
document processing, “taxonomy tagging” and more
• Microsoft’s Azure Custom Vision
o An image-recognition service for building, deploying and improving
image identifier models
• Tableau AI
o Data visualization tool
• Anyword
o AI-based marketing tool that can be applied to social media
• Sheeko
o Open source software to generate captions and keywords for images
• AgentGPT
o Custom AI agent that can be trained to plan and execute goal oriented
tasks, with wide potential application

Publishing
AI will have an impact on scholarly publishing in different ways, including the
production of OER resources. Publishers will need to create policies around
submissions that have been generated, in whole or in part, by generative AI.

At the same time, researchers and publishers will be able to streamline the publishing
process by using new tools for identifying appropriate journals for submission,
screening manuscripts or redirecting manuscripts to other journals, proofreading,
creating abstracts, formatting manuscripts, and more. Although a topic of discussion,
AI has yet to replace humans for the actual task of peer review and some funders
have banned AI for the purposes of their peer review processes; but this is a space to
be watched as developments are likely.

Examples of tools for editorial management and peer review

• iThenticate
o Identifies plagiarism, including copied text and non-idiomatic translated
text
• Edanz Journal Selector
o Recommends journals by comparing abstract text
• EndNote Manuscript Matcher
o Compares Web of Science data against title, abstract and references

CANADIAN ASSOCIATION OF RESEARCH LIBRARIES  BIBLIOTHÈQUES DE RECHERCHE DU CANADA 8

 • Penelope.ai
o Verifies formatting against journal requirements

Personal work / professional competency

AI tools can improve productivity for a variety of tasks, such as drafting emails, press
releases, and event posters. For example, Microsoft's AI companion, Copilot, will span
across all of Microsoft's applications, including Microsoft 365, Windows 11, Edge, and
Bing, so users will be able to get AI assistance across their workflow using MS
products.

Examples of task-based tools

• Grammarly
o Writing assistance
• InterviewMe AI
o Practice interviews for software engineers
• Rationale
o For decision making, including pros & cons analysis, SWOT analysis,
multi-option analysis, and outcomes analysis
• Taskade
o Automate workflows

AI tools can be used to automatically translate text into a variety of languages. In

addition to the utility machine translation provides to operations and research,
language tools may help level the playing field for researchers, students and library
workers who lack competency but need to write and publish in English or French. AI is
contributing to the revitalization of Indigenous languages, though existing AI
platforms still fall short when it comes to recognizing Indigenous language and
produce a high volume of hallucinations. Libraries should be aware of issues around
data sovereignty when engaging with AI in the context of Indigenous languages and
culture and copyright issues with translation activities, more broadly.

• Reverso
o AI based text translation in over 15 languages; additional services include
full document translation, integrated grammar checks and dictionaries,
and conjugations tools
• Deepl
o Online translation service that can be integrated directly into products
and platforms
• Tradooit
o AI based translation service adapted for Canadian vocabulary

CANADIAN ASSOCIATION OF RESEARCH LIBRARIES  BIBLIOTHÈQUES DE RECHERCHE DU CANADA 9

 • Google Translate
o The ubiquitous translation tool is especially useful for mobile users
• ReadAlong Studio
o Open-source software that uses elements of AI to create read-alongs in
27 Indigenous languages

Policy
As governments consider the broad, societal implications of AI as a transformational
technology, our views and expertise can contribute to local, provincial, and federal
policy development. Our voice needs to embrace the new technologies, while also
suggesting how risks might best be mitigated.

Institutional policies to address ethical and legal considerations are emerging rapidly.
They may address many of the areas mentioned here, but tend to be focused on
academic integrity, disclosure and citation. Addressing these issues in a rapidly
developing landscape will require ongoing research and attention.

The Government of Canada recently released a Voluntary Code of Conduct on the

Responsible Development and Management of Advanced Generative AI Systems, and
has tabled Bill C-27, Artificial Intelligence and Data Act (AIDA). Such instruments aim
to address the concerns below.

Misinformation
Misinformation includes both unintentional misinformation (e.g., hallucinations) and
intentional disinformation, with a range of consequences. For example, current AI
tools can make up citations to works that do not exist, necessitating vigilance and
vetting. Countering pernicious disinformation continues to be challenging despite the
development of tools to detect it. Human intervention remains key: understanding
effective prompt engineering (and training others) mitigates hallucinations, just as
vigilance and digital literacy effectively counters mis- and disinformation and other
harmful outputs.

Nature of inputs
Can published collections be data mined to support the production of LLMs? CORE-
GPT has done this based on open access articles. Working with materials that are
either behind paywalls or include all rights reserved content would expand the
knowledge base significantly and perhaps mitigate hallucinations in generated
content, but there could be significant legal implications.

CANADIAN ASSOCIATION OF RESEARCH LIBRARIES  BIBLIOTHÈQUES DE RECHERCHE DU CANADA 10

Copyright
At this point, AI-generated works do not appear to be eligible for copyright
protection in Canada and the availability of exceptions for the use of copyright-
protected content by AI tools is unclear, but the matter will likely be considered by
parliaments and brought repeatedly to the courts in the coming years. Libraries
should reinforce their reputations as copyright leaders on campus by staying up to
date on the situation and be aware of developments and decisions in other
jurisdictions. CARL submitted a response to the Consultation on a Modern Framework
for Artificial Intelligence and the Internet of Things in (2021) that lays out some of the
AI copyright concerns in terms of authorship, text and data mining and TPMs and the
Right to Repair.

Privacy
Many new tools make use of personal data, whether for training an AI system or to
check student papers for plagiarism. Libraries should be contributing to conversations
about appropriate use of student and researcher data and in what situations
individuals should be able to opt out of data collection.

While institutional research and ethics boards (REBs) provide oversight in terms of
the appropriate collection and storage of data acquired through research projects, as
issues around AI and privacy increase in complexity, libraries should be prepared to
support REBs in the ethical assessment of projects involving AI.

Bias
Current AI systems are known for being biased, since they have been trained on
biased data. They can also result in reinforcing existing biases of users. Libraries
should anticipate situations in which these biases could be harmful and create
strategies for mitigation.

Social impacts
Costs for some generative tools (e.g. premium editions) mean that some students –
and institutions – have limited access to certain AI tools, creating an equity issue.
Students who make use of AI-generated information, legitimately or dishonestly, may
be advantaged over those without skills, access, or inclination to do so. The increase
in automation may accelerate job loss and the exploitation of precarious labour. The
enormous consumption of electricity and carbon dioxide emissions associated with AI
technologies create a substantial environmental impact; for example, the increasingly
large datasets required to train LLMs require vast amounts of energy and create
substantial emissions.

CANADIAN ASSOCIATION OF RESEARCH LIBRARIES  BIBLIOTHÈQUES DE RECHERCHE DU CANADA 11

Conclusion
AI technologies offer considerable advantages to libraries, and moreover, the clock
cannot be turned back. The range of technologies, their uses, and relative risks and
benefits mean that they need to be considered on a case-by-case basis. Given the
number of systems currently at play in the library (and university) environment,
integration of new ones can be challenging, slow-paced, and resource-intensive;
institutional agility and responsiveness is key to staying current with technological
advances.

Aligning with institutional responses, libraries may wish to adopt local principles to
guide their use of AI technologies. These principles could address:

• User privacy and data security

• Algorithmic bias
• Investment and sustainability costs
• Training for library staff
• Equity and fairness
• Transparency with respect to AI systems in place
• Understanding the limitations of AI and the need for human oversight
• Text and data mining of library resources
• Regular risk assessment, monitoring and evaluation of systems and
practices to ensure safety and efficacy
• Collaborating and sharing best practices within the academic research
library community
• Advocating for and promoting responsible use of AI

With effort to keep abreast of developments and to thoughtfully seek out

opportunities and mitigate risks, libraries are well-positioned to deploy AI
technologies effectively and responsibly and cultivate strong, appropriate literacies
among their users. CARL will continue to monitor developments and consider whether
more issue-specific briefings would be useful to our members and the academic and
research library community.

Further reading
Policy and regulation
• Guidance for Generative AI in Education and Research (UNESCO)

CANADIAN ASSOCIATION OF RESEARCH LIBRARIES  BIBLIOTHÈQUES DE RECHERCHE DU CANADA 12

 • Voluntary Code of Conduct on the Responsible Development and
Management of Advanced Generative AI Systems (Innovation, Science and
Economic Development Canada)
• Statement on Generative AI (G7 Data Protection and Privacy Authorities)

How generative AI tools work

• Generative AI and Large Language models: Background and Contexts
• Generative AI, scholarly and cultural language models, and the return of
content

Discussions of academic library-related applications and

implications
• Generative AI and libraries: 7 contexts
• Looking towards a brighter future: the potentiality of AI and digital
transformations to library spaces
• ChatGPT: Implications for academic libraries
• Using Machine Learning for Finding Research
• Understanding AI Writing Tools and Their Uses for Teaching and Learning at
UC Berkeley
• AI Tools for Libraries
• ChatGPT-like AIs are coming to major science search engines (paywall)
• AI and human interactions in scholarly publishing
• AI and Scholarly Publishing: A View from Three Experts
• Artificial intelligence to support publishing and peer review: A summary and
review
• Can generative AI facilitate the research process?
• ChatGPT vs Bard vs Bing AI: The 2023 AI Battle
• The Prompt Engineering Librarian (paywall)
• AI Prompt Engineering Isn’t the Future
• Application of Artificial Intelligence (AI) In Libraries and Its Impact
• on Library Operations Review
• How Well Does ChatGPT Handle Reference Inquiries? An Analysis Based on
Question Types and Question Complexities
• Evaluating Generative AI Tools: a checklist
• The Rise of AI: Implications and Applications of Artificial Intelligence in
Academic Libraries
• An AI Toolkit for Libraries
• Library Copyright Alliance Principles for Copyright and Artificial Intelligence
• Developing a library strategic response to Artificial Intelligence

CANADIAN ASSOCIATION OF RESEARCH LIBRARIES  BIBLIOTHÈQUES DE RECHERCHE DU CANADA 13

CARL responses to artificial intelligence issues
• CARL submission to Pre-Budget consultations 2024 (2023)
• Consultation on a Modern Framework for Artificial Intelligence and the
Internet of Things. (2021)
• Brief to the Ontario Government’s consultation Trustworthy Artificial
Intelligence (AI) Framework. (2021)

Examples of AI LibGuides, Primers, and Other Resources in CARL

Institutions
University of Victoria:
• Scholarly use of A.I. tools
https://libguides.uvic.ca/AI_Tools
• The use of artificial intelligence tools and implications for Academic Integrity
https://teachanywhere.uvic.ca/academic-integrity/the-use-of-artificial-
intelligence-tools-and-implications-for-academic-integrity/
• UVic AI Safety Reading Group
https://uvicai.ca/reading-group/

Université du Québec à Montréal:

• ChatGPT et intelligence artificielle : utilisation de ChatGPT en contexte
universitaire
https://uqam-ca.libguides.com/ChatGPT_et_IA/Utilisations
• Intégrité académique et citer ChatGPT
• https://uqam-ca.libguides.com/ChatGPT_et_IA/Integrite_et_citer
• Forces et faiblesses
https://uqam-ca.libguides.com/ChatGPT_et_IA/Forces_faiblesses
• Invites efficaces
https://uqam-ca.libguides.com/ChatGPT_et_IA/Invites_efficaces

Carleton University:
• Artificial Intelligence (AI) - Tools
https://library.carleton.ca/guides/subject/artificial-intelligence-ai-tools
• Teaching Resources: Generative Artificial Intelligence
https://carleton.ca/tls/teachingresources/generative-artificial-intelligence/

CANADIAN ASSOCIATION OF RESEARCH LIBRARIES  BIBLIOTHÈQUES DE RECHERCHE DU CANADA 14