Research outcomes of exchange on current challenges in relation to Recent AI Breakthroughs and their Influence on Human Aspects in Personalization

• User Studies: A challenging question extends on how to deal with user studies: "Replace users with AI"?

- Research on user modeling requires user studies and these are notoriously hard for a variety of reasons (see UMAP proceedings and practices during the years)

• Large Language Models (LLMs): There are concerns and opportunities of LLMs in "human-in-the-loop" research, that relate to

- Generating synthetic data that do not reflect real users

- Data labeling with the support of LLMs maybe be proved untransparent and untruthful

- Crowdsourced users using LLMs to fill-in surveys

• Redefining the User: In this new reality, what is a user? 

- Are we referring to an "augmented user" = human + LLM

- A user should learn 21st century skills: critical thinking, posing questions (prompting)

• Potential: Advantages or opportunities may relate to

- No need to explain to a LLM the domain knowledge

- Preference elicitation can be more efficient

- Encouragement to define/ frame better the problem