Despite the significant leaps in the quality of online interaction in making us feel we’re in the same room, communicating in a ‘natural’ manner, one concern seems to be gaining traction as we move towards our fourth month of lockdown: how will this change face-to-face interaction IRL? It’s a crucial question, and one that’s hard to answer, given that video-calls allow us to sit facing each other, interacting in a way that by now feels habitual but that contains subtle (yet fundamental differences).
Of all of them, the one that has struck me most is the inability (due to the positioning of our webcams) to hold eye contact. This seemingly obvious flaw was first brought to my attention by Rana el Kaliouby, CEO of Affectiva and author of Girl Decoded, through an interview with Wired.
Throughout the book, it’s clear that for Rana el Kaliouby, the importance of emotional intelligence, or EQ, within technological innovation has been fundamental to her vision for decades. Yet her insights into how our EQ is affected when we interact in the digital world feels invaluable at a moment in history when an even greater amount of our social interactions are moving online. Published this year, but written pre covid-19, Girl Decoded is both a deeply personal memoir and an invaluable study into the importance of EQ for AI systems.
What makes this work so insightful is the melding of el Kaliouby’s personal quest to better understand her own emotions, whilst simultaneously mapping the effects that digital interactions have on our EQs. To achieve success in humanising technology, the coder herself has to understand herself by ‘turning on the spotlight’ on her own emotions. As el Kaliouby puts it: ‘my work and my personal story are inseparable; each flows into the other. And so this book is a chronicle of that dual journey – the quest to equip machines with EQ and, in the process, unlock my own EQ.’
Affectiva – Human Perception AI
Whilst there is a lengthy list thoroughly mapping out all of Affectiva’s core principles, one of the fundamental elements of how they work is patent throughout Girl Decoded: whether it be in equipping automotives or analytical media with affective AI or creating AI to assist autistic people in better understanding human emotions, Affectiva are committed to the ethical development and deployment of AI. Given the potential room for interpretation of what ethical deployment means, It would perhaps be easy for any AI company to give this as a guiding principle.
Yet in Affectiva’s case, we have clear evidence that principles really do come before profit. When Affectiva was first starting out, the national security branch of the United States government saw how useful their software would be for surveillance purposes. As a young start-up trying to get off their feet, el Kaliouby notes that turning down $40 million of investment felt almost suicidal. But admirably the team stuck to their core values of opt-in and data privacy, rejecting that their software be used for surveillance.
Within their push to create AI that is emotionally intelligent, one of the key premises of el Kaliouby’s work is that ‘an algorithm is only as open-minded and smart as the human that input it.’ As she documents throughout the book, inbuilt biases within algorithms can not only lead to technologies that exclude or even offend different ethnic groups – we’ve all heard of Google Photos’ algorithm classing people of colour as gorillas – but that render the technology futile.
Recognising that removing these biases isn’t straightforward, el Kaliouby notes how the early versions of her facial analysis software developed at MIT Media Lab didn’t work in China because different cultures (in this case one that prizes collective achievement over individual) require different emotional behaviour from their citizens. What’s most impressive is that, rather than simply extolling the achievements of Affectiva, el Kaliouby emphasises that issues with underlying bias within algorithmic software are ongoing. What is required is a commitment to perpetual improvement, rather than crossing a set finishing line.
Perhaps one of the most fascinating (and relevant) discoveries of el Kaliouby’s research is the ways in which digital interactions can render all of us functionally autistic. For those that suffer with autism, not only are different facial expressions and tone of voice confusing and frequently unintelligible, the very act of looking someone in the eye is often avoided due to the anxiety it induces. Whilst el Kaliouby isn’t trying to suggest that digital interactions are making us autistic, she neatly demonstrates how many of the ways we communicate, such as texting, render us functionally autistic as we simply have no access to facial expressions, tone of voice or body language.
Words establish meaning through their definitions alone; a way of communication which is different in essence from how we decipher information during face to face communication. So whilst Affectiva’s Brain Power device has made great strides in equipping autistic children with tools to ameliorate their communication difficulties, the application of such software opens many possibilities for improving the ways we all communicate online.
Socially desirable technology
Part of what el Kaliouby draws our attention to is that machines are often out of sync with our expectations. We expect that our machines should understand our frustration when they malfunction. We also expect that we should be understood when communicating our ideas through digital interfaces. Unfortunately, as it stands, the reality for most of us is quite far off.
Not only do we show less empathy towards others when interacting digitally, but the very devices we use to do so have little if no understanding of how we ourselves are feeling. In this sense, our contemporary technologies, whilst light-years ahead of us in terms of computation abilities, are also functionally autistic. Wouldn’t society be healthier if both we and our technologies had higher EQ?
Imagine that when communicating online, your interface could tell you in real time exactly how your interlocutor was feeling? Wouldn’t AI systems equipped with EQ that provided high-quality, personalised learning tools for children and adults be a societal benefit? Wouldn’t children benefit from learning assistants that helped them deal with emotionally challenging situations?
This handful of ideas, picked from the numerous that Affectiva are working with and developing, not only demonstrate how as a company they are leading the way in emotionally intelligent AI, but also shed light on the importance of humanising technology not just for the few, but for all of us as we move deeper into a world in which so many of our interactions are moving into the digital space. Now more than ever, el Kaliouby’s push to humanise technology brings to the forefront some of the most important questions we all face in considering how AI, for better or for worse, will shape our future societies.