Sven Laqua, PhD FRSA’s Post

View profile for Sven Laqua, PhD FRSA

Head of Design & Research, Director & Founder at Laqua Consulting specializing in human-centered design and UX strategy.

The UK Research and Innovation's new policy document on how to use AI tools responsibly is probably much more widely appropriate than the intended research funding application process. Some highlights: - Never input sensitive or personal information (without consent) - Consider the risk of bias - Ensure outputs are not falsified, fabricated, plagiarised, misrepresented I also find the assessor angle on not using AI tools to 'outsource' personal evaluation and judgement very relevant but likely hard to apply in practice in many scenarios. One of the core use cases at the moment for LLMs like ChatGPT is summarising meetings, long documents, etc. Without validating the accurateness and completeness of the summary, "AI judgement" is very quickly baked into any subsequent human decision making. #AI #Policy #Ethics #HumanFactors #DecisionMaking The full policy is here: https://lnkd.in/dDD6B3mC

Use of generative artificial intelligence in application preparation and assessment

Use of generative artificial intelligence in application preparation and assessment

ukri.org

Godwin Josh

Co-Founder of Altrosyn and DIrector at CDTECH | Inventor | Manufacturer

5mo

The real challenge lies in defining "accuracy" and "completeness" for summaries, especially when dealing with nuanced or subjective content. LLMs often excel at capturing surface-level information but struggle with deeper contextual understanding. This can lead to biased or incomplete summaries that inadvertently influence human decision-making. You talked about the assessor angle on not using AI tools to 'outsource' personal evaluation and judgement very relevant but likely hard to apply in practice in many scenarios. Imagine a scenario where an LLM is tasked with summarizing legal documents for a court case, how would you technically use the concept of "accuracy" and "completeness" to ensure the generated summary captures all relevant legal nuances and avoids potential misrepresentation?

Like
Reply

To view or add a comment, sign in

Explore topics