AI search tools
AI tools are changing the academic information landscape – this also affects scientific searching for reliable information.
Licensed tools
Consensus access and account
- Premium access for all members of the University of Basel
- Access – even from outside the campus network – is effected by logging in with your Unibas email address.
- Scite: https://scite.ai/ Academic literature search in pending and proprietary content
- Elicit: https://elicit.com Academic literature search based on Semantic Scholar
- ScopusAI: https://www.elsevier.com/products/scopus/scopus-ai Academic literature search in Scopus
- Web of Science Assistant: academic literature search in Web of Science
- Research Rabbit: https://www.researchrabbit.ai/ Visualization tool for connections in the research literature
- Open Knowledge Maps: https://openknowledgemaps.org/ Visualization tool for connections in the research literature
External links: AI resources:
- Website of the Virtual Center of Excellence Teaching and Learning Writing with AI at Kiel UAS: www.vkkiwa.de/ki-ressourcen/ki-tools/
- Georgetown University website: https://guides.library.georgetown.edu/ai/tools
Use of AI for academic searching for information
Opportunities and risks
- Opportunities lie in increasing efficiency, discovering new correlations and managing large amounts of data.
- Risks include potential bias as a result of faulty training data, "hallucinations" of false information and a possible dependency that may impair critical thinking.
A conscious and critical approach to AI tools is essential. The results of an AI-based literature search should be scrutinised and if necessary, compared with other search methods.
Tips for working with AI tools
- Review sources: Some tools create references that look real but do not actually exist.
- Critical handling of search results: AI tools are based on probabilities and therefore make mistakes. For this reason, checking with other tools or through your own searches is essential.
- Which sources are considered in the search? Each tool searches a specific collection of literature or specific databases. It is therefore possible that important sources are not taken into account and have to be searched for in a different way.
- AI bias: The statements made by AI tools are not neutral, but may contain biases due to the training data or the design of the tools. Possible distortions should therefore be taken into account.
Numerous AI tools are currently available for information searching. There are significant differences between the tools, which must be taken into account.
Here are a few distinguishing features:
- Up-to-dateness: is the answer generated using only static training data or also using current search results?
- Which data is used for the search? (e.g. open databases such as Semantic Scholar, proprietary databases, abstracts or full texts of articles, Google web search, etc.)
- Are the sources cited in the answer?
It is important when searching for academic information, that the results are generated from academic and reliable sources and include references.
Many tools use Retrieval Augmented Generation (RAG) to extract information from selected documents. Academic publications are searched for as an answer to a user question using scientific search engines and in specialist databases and used as sources for generating the answer. In addition to these conceptual differences, there are of course also differences in quality, i.e. the quality of the answers varies greatly.