Ethics of AI algorithms: looking into PubMed’s “Best Match” Algorithm
Artificial intelligence is in our everyday lives; whether it’s through our google searches, Netflix recommendations, or self-driven cars; it is nudging our behaviours or opinions in a certain direction. Discussions about how algorithms reinforce stereotypes or how biases are included in the programming of algorithms are essential to bring awareness and limit the repercussions on people and society.
Where do library systems and resources fit in all this? We believe it’s important to investigate the ethical implications of artificial intelligence in library systems and databases, especially as these search engines are adding artificial intelligence layers. PubMed, a freely available health science database, recently updated its default ranking algorithm to LambdaMART, a Learning-to-Rank algorithm that uses machine learning to improve relevancy. This is great on many levels, especially when we look into user search behaviours and user expectations of such systems. However, it also comes with many ethical implications and concerns about how it might affect researchers, clinicians, and research outcomes. This presentation will, first, explain how PubMed uses artificial intelligence in the Best Match algorithm. We will then outline possible ethical implications of artificial intelligence in PubMed and how it might impact a wide range of users.
