Skip to Main Content

History: AI and the Library

Purpose: This library guide offers an overview of how you can utilize Generative Artificial Intelligence (AI) to enhance your use of library resources at Queen’s University Belfast.

Remember, before incorporating Generative AI into your research or writing, ensure you:

After completing these steps, explore this guide for advice on using AI effectively and ethically in literature searches and in sourcing materials for your studies.

Key Message: Generative AI can help expedite some elements of traditional library research, but it should not be used as an alternative resource for finding peer-reviewed, academic sources. Always validate information using the Library's academic databases and its wider range of resources. 

The Role of Generative AI in Academic Research

What is Generative AI?

Generative AI tools, such as ChatGPT, are advanced systems that create text based on patterns learned from vast amounts of data. These tools are trained on large datasets that include books, articles, websites, and other publicly available information. By recognizing linguistic structures and common themes, they can produce coherent and relevant responses to a wide range of queries.

However, it's important to note that while these tools may provide helpful, contextually accurate information, they do not access real-time data or all peer-reviewed academic sources. They cannot search university library databases nor Google Scholar, meaning that the information they generate is based on past data and not guaranteed to be academically verified or up-to-date. Therefore, when using AI for academic purposes, it's essential to cross-reference with credible sources.

Limitations of Generative AI:

  • Generative AI tools do not have access to real-time data or live information updates.
  • The content generated is based on past data, not guaranteed to be academically verified or current, making it essential to cross-check content with credible, authoritative sources for academic use.
  • The models do not have access to subscription-based academic databases like Scopus, Web of Science, or PubMed.
  • Outputs are not always accurate or up to date; they lack the rigorous vetting that peer-reviewed articles undergo.

The QUB AI Hub provides further resources and official guidelines on AI use in academic work for students and academic staff.  

When and How to Use Gen AI models in Literature Searches

Complementing Traditional Research:

There are two elements to consider: AI models used as a tool or for content. It's important to recognize that while generative AI tools can support your research activities, they cannot replace traditional academic databases. (See section on Role of Gen AI & Academic Research). Relying on The Library at QUB for comprehensive and credible sources should remain central to your research and searching processes.

Effective Uses of Generative AI:

1. Keyword generation:

Generative AI tools can save time by helping you discover and select key terms to create search strings, which are then used to search library databases comprehensively.

  • Asking for Key Terms: Request an AI tool e.g. Chat GPT or Claude, to suggest key terms or synonyms based on your assignment title or research question. For example, if your topic is "sustainable urban development," you can ask the tool to provide related keywords or phrases.

  • Exploring Synonyms and Variants: Ask the AI model to generate different synonyms or variations of important terms in your research. This helps broaden your search and ensures you cover various aspects of your topic.

  • Refining Your Search Terms: Discuss your research focus with the AI model to get a list of relevant terms that might not be immediately obvious. For instance, you might ask, "What are some related terms to 'climate resilience'?"

  • Generating Related Concepts: Ask the AI tool to provide related concepts or subtopics that could help in expanding or refining your literature search.

Practical Tip: Refine your search terms with more specific questions or prompts for more focused results.

2. Formulating Research Questions:

AI can assist in breaking down complex research questions into simpler components. This is especially useful in the early stages of structuring a research query.

3. Exploring Topics and Definitions:

Use an AI model to get a basic overview of a topic or to clarify definitions before diving into scholarly sources. However, always verify the information with the Library's authoritative sources and be mindful of hallucinations. (See section Evaluating AI-Generated Information).

Practical Tip: Keep a note of the more successful prompts you use and create your own 'prompt library'.

Examples of ChatGPT prompts:

Example of prompts or questions you can use to source keywords for a search strategy include:

  • "What are some key terms to use in a literature search on renewable energy policies in Europe?"
  • "Can you suggest some themes to explore in research on mental health services for students?"

AI and Library Databases & Resources

How do library databases use AI?

AI has long enhanced search performance in library bibliographic, full-text, and other databases by enabling users to control searches with keywords and filters, while the AI efficiently processes and delivers relevant results.

With advances in generative and conversational AI, we relinquish up some control over the search process, relying on AI to deliver relevant results. As we interact more, the AI learns and improves its accuracy. Conversational AI allows natural language searches, shifting away from traditional database methods. The University's guidance on this is detailed in the Prompt Engineering section.

Generative AI models are impressive tools that continue to improve, but their content is not guaranteed to be academically verified. While they may include free or open-access research, they cannot access content behind paywalls or within subscription services. Therefore, library databases, which are paid for and accessible to all QUB staff and students, remain essential for finding peer-reviewed and academic materials.

Note: Library database suppliers and publishers are beginning to offer conversational AI interfaces and the Library will provide progress updates as they are implemented and launched. 

Importance of Databases:

  • Generative AI tools cannot access peer-reviewed sources like those found in academic databases.
  • For credible, academically verified information, it is crucial to use the scholarly databases provided by the Library.
  • Always cross-reference AI-generated content with trusted sources to ensure academic accuracy.

Key Library Resources:

AI Tools for Literature Searching:

A number of AI tools that focus more on academic content have been and are being developed. Examples include  PerplexityAIConsensus, and Elicit. While each operates differently, they can all help in offering a general overview of a research landscape. However, these tools should not replace the critical task of reading and thinking about the literature yourself, particularly in relation to your specific research questions and priorities. The University's guide on Using AI provides more information about these research tools.

Remember, generative AI tools typically cannot access any academic sources which require a subscription, so using only these tools will exclude a significant proportion of the research literature you will need in your studies.

Integrating AI-Generated Keywords into Database Searches

Using AI-Suggested Keywords for Literature Searching

If you've used an AI tool to generate keywords (see When and How to Use Gen AI Models in Literature Searches), you can incorporate these into a search strategy for more effective use of the Library's databases.

Examples Using ChatGPT:

1. Single Question Approach
Ask ChatGPT, "What are key terms used for 'sustainable urban development'?" From the suggestions, you can choose the most relevant keywords. For example, ChatGPT might suggest terms like green infrastructure, urban resilience, sustainable transportation, and urban regeneration. Searching each term individually in an academic database will help you find peer-reviewed literature on each topic.

2. Two-Part Query Approach
Alternatively, you can ask two targeted questions for a more refined search: "What are alternative words for sustainability?" and "What are alternative terms for urban development?" Use the results to build search strings, combining alternative terms with OR and concepts with AND for a comprehensive search in the Library's databases.
(sustainability OR resilience...) AND (urban development OR urban renewal...)

  • Sustainability keywords:
    sustainability OR resilience OR eco-friendliness OR environmental responsibility OR durability OR regeneration OR conservation OR future proofing OR carbon neutral

  • Urban development keywords:
    urban development OR urban renewal OR urban expansion OR town planning OR urban regeneration

Further Assistance
These search strings can be further refined for database use with techniques like truncation and proximity searching. Truncation is a search technique using a symbol (often an asterisk '*' ) at the end of a word stem to find all variations. For example, "educat*" retrieves "education," "educator," "educating," etc., broadening search results. Proximity searching retrieves results where two or more terms appear within a specified number of words, improving relevance.

For more help on developing your search strategy, visit your Subject Guide.

Evaluating AI-Generated Information

The Importance of Critical Evaluation:

All content from AI tools needs to be critically evaluated. AI can provide useful insights, but it's essential to fact-check and cross-reference information with reliable academic sources to ensure its accuracy and credibility.

Spotting Errors:

To identify misinformation or inaccuracies in AI-generated content:

  • Watch for vague or unclear claims.
  • Check for outdated information.
  • Look for a lack of proper citations.
  • Be aware that Generative AI can present inaccurate or made up information 
    • AI "hallucinations" refer to instances where an AI model generates information that is inaccurate, fabricated, or nonsensical, despite being presented as plausible or authoritative. These errors can occur due to the AI's reliance on patterns in data rather than actual understanding or reasoning. Historical inaccuracies, fabricated facts and misattributions are all examples of types of AI hallucinations. The article by Farquhar et. al. published in Nature, June 2024 by explains more about "Detecting hallucinations in large language models using semantic entropy" 
    • Be especially cautious when considering any citations to research literature produced by Gen AI - frequently, the sources cited do not actually exist.

Best Practices:

  • Double-check any facts or references provided by AI tools in trusted academic sources.
  • Always cite peer-reviewed sources in academic work, not AI tools themselves.

Ethical Considerations in Using Generative AI

Academic Integrity:

The University provides guidance on AI and integrity and academic rigour, and assessment and feedback along with a statement in relation to AI and Research.

Citing AI:

Please see the Citing and Referencing AI section of this guide. 

Privacy & Data Concerns:

The University provides guidance on using AI responsibly which includes advice on Privacy and Security

Staff and students are encouraged not to input sensitive information or personal details when using AI tools and models, as AI platforms may store or process this data.

Further information is available from:

JISC - A pathway towards responsible, ethical AI

The Alan Turing Institute Artificial intelligence (Safe and Ethical)