#WebSci20 – Paper Session 4: AI History and Trust by Anna Kent Muller
And in our knowledge infrastructure today… the social machine that is video awards! @pipwillcox @OxfordeResearch @UkNatArcRes #WebSci20 pic.twitter.com/EpfspvdR7o
— David De Roure (@dder) July 10, 2020
Posted on behalf of Anna Kent Muller
This session featured four papers around the concept of the history of AI, social machines and trust. We heard two papers (one by Zhou Shao and the other Sha Yuan) Â discuss academic knowledge, and the evolution of the discipline. These raised questions around how we ensure the field is inclusive, how we continue to promote and match the trend of international collaboration whilst protecting against any dominance from certain countries of companies. General agreement falls on how we need to see the evolution of a discipline that involves diverse and spread actors, and make sure that academic knowledge is not concentrated on a few.
The two remaining papers covered trust and social machines. The first by Harini Suresh detailed a study that looked at the role of human interaction in Machine Learning. She argued for us to consider human interaction in our development of AI and ML systems, specifically I liked her acknowledgement that neither humans nor AI are perfect, utilising them together enables us to protect against the areas they mess up, as usually these are different. The discussion following her presentation interestingly revolved around how humans might need to know more than just the outcome of machine learning, how understanding the decision making process of machine learning could help us to make better decisions and rely upon ML models in a stronger and more collaborative way. Similarly, how do we explain to individuals the areas of weakness within a model, and allow them to use that within their decision making process.
David De Roure and Pip Willcox’s paper on social machines discussed the role of knowledge infrastructures in research. I liked their comparison between Web Science and how we have stepped back to look at the web, with what needs to be done to look at social machines. Social machines innately involve us, in the same way as the web involves us – they’re both socially constructed. The methodological approach discussed in reference to the life of a physical object (in their example Bodleian First Folio of Shakespeare’s Plays), raised interesting questions around discussing the scholarly primitives (as defined by John Unsworth) relevant to social machines – importantly they added co-creating, collaborative and contributory as an important scholarly primitive of a social machine. This is our opportunity as people, now to be authors and creators of social infrastructures, to be mindful of the infrastructure we are creating.I liked how they rounded together the conference so far, by highlighting how AI has the ability to amplify all voices.