Professor Alan Smeaton:

I am up next. I thank Senator Ruane for her great question. She asked a computer scientist a computer science question and the Cathaoirleach is only allowing a minute for an answer.

There was a lot of coverage that AI systems, which we used in recommender systems, were biased. We have seen lots of examples. I think Professor Ahern knows the great example of an AI system recommending whether somebody should get parole or not and the AI system was blamed for that. However, it is not the AI system that should be blamed; it is the training data that reveals the bias. In one sense, from a societal point of view, it is a real positive that these biases inherent in the behaviour of people making decisions is finally revealed. As a computer scientist, it presents a wonderful technical problem in that the cases are then synthesised or balanced in order that they even out and that the future output of recommender systems is less biased. Raising the awareness of the bias and the existence of that in the first place is a great positive.

Plug me in and I will say exactly what Dr. Scanlon said. Once a person knows and has a familiarity with AI literacy, and he or she knows his or her social media feed is chosen by an algorithm, and how that algorithm is biasing the feed to get a more emotive reaction and more clicks, that is great. It is a fundamental and a real positive.