Lightnews — Scholar-powered news

David Haynes @jdavidhaynes.bsky.social · 4d

Thanks to Aida Slavic and #ISKOUK for organising a brilliant biennial conference on 'Knowledge Organization in the Age of AI'

2

David Haynes @jdavidhaynes.bsky.social · 4d

Maybe we cannot 'solve' the problem of different representations of communities. It is more of a process (e.g. by participatory governance). Otherwise we could be guilty of erasure of other experience. It is difficult and could be uncomfortable. Joseph Busch at the #ISKOUK 2025 Conference.

1 1

David Haynes @jdavidhaynes.bsky.social · 4d

Terminology is used in specific context. Participatory governance - include the communities that are being described by those systems - to more fairly represent communities. Julaine Clunis at the #ISKOUK 2025 Conference

2

David Haynes @jdavidhaynes.bsky.social · 4d

Joane Casenave, Widad M El Hadi, and Thibault Grison from the University of Lille (France)  are doing a joint presentation on 'Artificial intelligence, KO and information mediation on the Web: from ethical dimensions to social responsibility' at the #ISKOUK 2025 Conference

1

David Haynes @jdavidhaynes.bsky.social · 4d

Good to see a typology of explainable AI (XAI) methods used for analysing KOSs in Julaine Clunis' paper at the #ISKOUK 2025 Conference

1

David Haynes @jdavidhaynes.bsky.social · 4d

Very interested to hear about Biases in KOSs in Julaine Clunis' talk at the #ISKOUK 2025 Conference. This is a real problem when it comes to algorithmic bias.

1

David Haynes @jdavidhaynes.bsky.social · 4d

Need a formal evaluation of 2dseach rather than the error checking and anecdotal evaluations - according to Tony Russell-Rose at #ISKOUK 2025 Conference

David Haynes @jdavidhaynes.bsky.social · 4d

Refinement of searches through exploration built into the LEGO model described by Tony Russell-Rose at the #ISKOUK 2025 Conference.

David Haynes @jdavidhaynes.bsky.social · 4d

What happens when you try to build a systematic search system 'for real'? What are the practicalities? Prof. Tony Russell-Rose at #ISKOUK 2025 Conference

David Haynes @jdavidhaynes.bsky.social · 4d

Tony Russell-Rose of City St George's, University of London (UK)  on 'Building and deploying LLMs for search and retrieval' at the #ISKOUK 2025 Conference

1

David Haynes @jdavidhaynes.bsky.social · 4d

In open world AI only 20 or 30 ontologies are used out of nearly 900 ontologies reviewed by Mayukh Bagchi and presented at the #ISKOUK 2025 Conference

1

David Haynes @jdavidhaynes.bsky.social · 4d

Mayukh Bagchi of the University of Trento (Italy)  talks about 'Generative knowledge organization via human-LLM collaboration' at the #ISKOUK 2025 Conference

David Haynes @jdavidhaynes.bsky.social · 4d

Even small errors in automatic generation of a KOS can have a disproportionate effect on its performance. @angelosalatino.bsky.social #ISKOUK 2025 Conference

1 2

David Haynes @jdavidhaynes.bsky.social · 5d

Using AI to generate KOSs - @angelosalatino.bsky.social proposes an alternative to statistical inference for automatic generation of ontologies of research topics #ISKOUK 2025 Conference

1 1

David Haynes @jdavidhaynes.bsky.social · 5d

Machine-readable thesauri are essential. @angelosalatino.bsky.social identifies some major thesauri are only available as PDFs. There are a good number available in RDF. #ISKOUK 2025 conference

1 1

David Haynes @jdavidhaynes.bsky.social · 5d

Angelo Salatino @angelosalatino.bsky.social of The Open University (UK)  talks about 'Knowledge organization systems of research fields: overview and automatic generation' at the #ISKOUK 2025 Conference

1 2

David Haynes @jdavidhaynes.bsky.social · 5d

Tanja Svarre @drsvarre.bsky.social opens the third session of the #ISKOUK 2025 Conference with a presentation by Angelo Salatino
@angelosalatino.bsky.social.

1 2

David Haynes @jdavidhaynes.bsky.social · 5d

Why don't we use KOSs in addition to LLMs? A question put to us by Joseph Busch at the #ISKOUK 2025 Conference

2

Reposted by David Haynes

Prof Sam Illingworth @samillingworth.com · 5d

🤖 Is AI ever really neutral?

Neutrality is an illusion. AI tools reflect the biases in their data and design.

This week’s Slow AI practice: after any polished AI answer, pause and ask, who might this disadvantage?

theslowai.substack.com/p/ai-neutral...

#SlowAI #AIethics #AI #SciComm

When AI Sounds Neutral, Is It?

AI is not neutral. It mirrors bias in data and design. Learn a simple practice to ask who is left out, and why diverse voices matter for real solutions.

theslowai.substack.com

2 3 8

David Haynes @jdavidhaynes.bsky.social · 5d

Should prompting engineering be a part of preparation of students for the workplace? Joseph Busch has introduced some interesting concepts in his overview of the development of AI in retrieval at the #ISKOUK 2025 conference

David Haynes @jdavidhaynes.bsky.social · 5d

Joseph Busch of Taxonomy Strategies (USA)  talks about 'The case for general purpose categorizers as part of the AI ecosystem' at the #ISKOUK 2025 conference

2

David Haynes @jdavidhaynes.bsky.social · 5d

Fascinating to learn about 'Potemkin Understanding' - i.e. a facade (or illusion) of understanding exhibited by LLMs - Ronald Siebes at the #ISKOUK 2025 conference

2

David Haynes @jdavidhaynes.bsky.social · 5d

Ronald Siebes from the Vrije Universiteit Amsterdam (Netherlands)  talks about 'Potemkin Understanding: the illusion of thinking' at the #ISKOUK 2025 Conference

1

David Haynes @jdavidhaynes.bsky.social · 5d

@gnoli.eu chaired the first session of the #ISKOUK 2025 Conference. An emerging theme seems to be about the use of AI to enhance indexing of resources, but with human oversight. We do not seem to be yet at the stage where AI generated indexing or metadata is totally reliable.

1 1

David Haynes @jdavidhaynes.bsky.social · 5d

Ziyoung Park from Hansung University (South Korea) talks about ' Co-creating with GenAI and expert review: empowering KOS projects through design, automation and recommendation' #ISKOUK 2025 Conference