AI for Humans: a personal agenda

French surveys have revealed that for one third of employees, professional digital tools are a cause of stress, and that for the general public, one in five adults never used digital tools or stopped using them because of their difficulty. Moreover, negative feelings are reported by more than 60% of the interviewees of a survey on Artificial Intelligence. A large literature is now devoted to the explanatory factors but the role of engineers and developers of digital technologies is still poorly studied. The popular adage "it's computers’ fault" captures this invisibility. Yet major international AI conferences now include sessions dedicated to ethical issues and societal impacts that engage scientists. In this presentation, which will take the liberty of a reflexive posture, I will analyze different avenues of research that I have explored in an attempt to develop data mining and artificial intelligence algorithms "more adapted" to human users.

Three approaches will be discussed:
(1) human-algorithm coupling through knowledge modeling with ontologies, and "relevant" information filtering with data visualization
(2) human-inspired algorithms based on usage observations
(3) machine learning algorithms accessibility by including available computing resources in the algorithm development.

Talk followed by an apéro to continue the discussion.

This talk is part of DH Research Seminar, a series of talks organised by the Digital Humanities Institute at the EPFL given by researchers from a wide range of backgrounds and aiming at presenting the vast array of subjects covered by Digital Humanities.

More events here: https://memento.epfl.ch/dhi/?period=365