My first bit of off duty reading on my return from an excellent Ariba Live (recordings here) was written by Rory Stewart – serial walker, accomplished writer, and outsider to become the UK’s next prime minister; in case you missed it, the current one’s been polished off by Brexit (perhaps something for discussion in Deloitte’s Brexit Briefing on 10th June, recording likely available). Stewart’s article ends with the cliff hanger Is it possible for a politician to talk about love? This summoned a smile from me, similar to when Rana el Kaliouby queried whether computers could be emotionally intelligent in her closing speech, at Ariba Live. In actuality, the query is less something to smile about, and more a matter for serious thought as signalled by el Kaliouby’s call for there to be a “New Social Contract between humans and AI: Based on Reciprocal Trust”.
Not a proposition I had ever considered before and quite something to wrap my head around. But then the point of going to an event like Live is to learn things, things which help you to think creatively and differently. Even getting started with this is tricky though. I’m not wholly sure I know what Trust is. In classic psychological terms, trust is considered more a ‘feeder’ of emotion. Decided a bit further it seems “most philosophers believe that trust is a kind of reliance on another person that, when violated, provokes not only disappointment but a feeling of betrayal“ i.e. broken reliance feeds the feeling of betrayal. And that’s just the tip of the iceberg but, as a practical alternative to taking up a phd in philosophy I think it’s better to leave it at the Bon Jovi view, that it’s another word for Love, and move on.
So what does AI need to do to journey towards being party to a new Social Contract. First step: the AI needs to recognise Trust and emotional recognition bringing us squarely back to el Kaliouby’s work. The foundations of Affective Computing were laid by Rosalind Picard who invented the term and with whom El Kaliouby founded Affectiva. Picard now focuses on medical, rather than business applications. Her recent presentation to the Brain Mind Summit is both short and a great watch, which she begins with revealing that she dreams of a future in which she is able to get her stress, mood and health forecast on her phone and moves on to explaining why she believes it’s both possible and important. Such bold ambitions, I think, serve to highlight, why affective computing is an easy target for melodramatic headlines even when linked to some otherwise measured and well informed reporting.
Back at Live el Kaliouby explained that the basis of Affectiva’s work and affective computing in general is the coding of varied facial expression data. She stressed Affectiva is meticulous in trying to ensure that data samples are taken from representative groups of people to ensure that algorithms don’t develop biased or skewed perceptions of any group of people. Once the computer can recognise the expressions the question is what real world benefits can be achieved from this. The possibilities do seem endless but currently testing for drowsiness in car drivers and gauging emotional changes of viewers as they watch advertisements are leading the line.