“Society will get the AI it deserves.” - Joscha Bach, What to Think About Machines That Think Today, AI-powered machines can defeat the most skilled players in chess, game of Go, Jeopardy and poker. They can outperform best doctors in diagnosing most complex health conditions. Yet when it comes to understanding a wide range of human emotions, AI is not very intuitive. In his 2016 Google talk, Professor David Gelernter said: “ Understanding emotion as an encoding and simulating function is one of the most important unsolved problems - in fact, ignored problems and missed opportunities in AI today.”  Why do we need to teach thinking machines to understand human emotions? While applications for emotionally intelligent AI are too many to list here, let me mention three areas that we, at Heartbeat AI, are most fascinated about.

Deeper understanding and better prediction of irrational human behavior

We don’t need to try hard to convince anyone today that human beings are emotional first, and rational second. Decisions that we make every minute - from what flavour of Starbucks to buy to who to vote for - are made deep inside our emotional minds influenced by many conscious reasons and sub-conscious “primes.” Last Summer, we decided to run an experiment and see if we could build a predictive algorithm for national elections using unstructured text data from surveys. Traditional election polling asks questions like “If elections were held today, who would you vote for?” or “Who do you think will become the next Prime Minister of Australia?”. We picked 3 traditional questions and added two open-ended questions asking how people feel about each candidate winning the elections. The survey ran on Google Survey Platform for 6 weeks prior to the June 2nd Australian elections and November 8th US elections. Then we used a predictive model to call the results. In both cases, emotional measurements helped improve the level of predictability, and therefore made correct predictions for Australia (predicted for Bill Shorten vs. Malcolm Turnbull: 44% vs 56%, actual: 48% vs. 52%) and US (Donald Trump vs. Hillary Clinton):

  • Florida: predicted 52% versus 48% actual: 51% vs. 49%;
  • Iowa: predicted 58% vs. 42%, actual 55% vs. 45%;
  • Ohio: predicted - 63% vs. 37%, actual - 54% vs. 46%;
  • Pennsylvania: predicted - 54% vs. 46%, actual - 51% vs. 49%

What was even more interesting is that the models showed what particular emotions drove voters’ behaviour, which explained the deep “why” behind people’s decisions.

Emotion text analytics is a new, possibly ground breaking, area of research and analytics. What if we could rely on text data to be the “secret sauce” for accurate prediction of future events that are based on human decisions, such as elections, consumer behaviour, public opinion and social movements.

Apps that will help people deal with stress, change and suffering

Eight years ago, I took a break from my career in market research and went to study psychotherapy. I had a small private practice for a couple of years, helping people deal with loss, trauma, distractive anger, and debilitative fear and anxiety. Inside my own family and friends, I witnessed hard struggles my loved ones fought with alcohol and drug addictions. At the end, I did not find the stamina to be a psychotherapist, but I never stopped wondering if good therapy could be scaled, free and available when and where people need it the most. Then two years ago, an idea of an algorithm that could recognize a wide range of emotions from language - written and spoken - came alive. That’s how Heartbeat was born.

This Spring, we are starting a pilot with a small residential drug and alcohol addiction rehab in Ontario. We will give people a chance to share how they feel in the most difficult moments of their lives, recognize and name their emotions, and track their progress. We hope to build predictive algorithms that could help prevent relapse and foster resilience.

This is just the first step in a long journey. With 6 billion people with access to mobile phones by 2020, apps can help us not only find a good restaurant or a Pokemon, but peace of mind too. In 2016, the Guardian stated that “AI could leave half of world unemployed”. Imagine the stress and anxiety this would trigger! AI-powered apps could emotionally support millions of people transition from losing a job to retraining for new careers, and building new lives.

Building the foundation for benevolent AI

“Artificial minds will be faster, more accurate, more alert, more aware and comprehensive than their human counterparts. AI will replace human decision makers, administrators, inventors, engineers, scientists, military strategists, designers, advertisers, and of course AI programmers,” writes Joscha Bach of MIT Media Labs. He continues: ” The motives of our artificial minds will (at least initially) be those of the organizations, corporations, groups and individuals that make use of their intelligence. If the business model of a company is not benevolent, than AI has the potential to make that company truly dangerous. Likewise, if an organization aims at improving the human condition, than AI might make that organization more efficient in realizing its benevolent potential.”

When something possesses that much power, its makers ought to carry a high level of responsibility. It’s absolutely paramount to ensure that our powerful machines have benevolent “souls”, including intentional kindness, deep understanding of human condition, and cognitive empathy. What if AI could help us solve the biggest Global challenges, help humanity create Global governance, social and corporate systems, and ultimately re-build trust?

Heartbeat is building a platform for people to share emotions and feelings about “hot” topics, burning political issues, the future and life in general. We want to use emotion text analytics to help people understand their own emotions, see emotions of others across the world, and with that, have a window into each other’s souls. I hope that a Global “empathy meter” will grow if instead of opinions and agendas, people share personal feelings and see how fear and anger drives us away from each other.

If we are honest with ourselves, we would admit that we get the partner and the government we deserve. At the end of the day, we will get the AI we deserve.

Lana Novikova is CEO & Founder of Heartbeat AI, which she created to teach machines Emotional Intelligence (EQ) in hope that one day machines can help people build their EQ. Today, Heartbeat has an award-winning SaaS that helps marketers around the world understand consumer emotions. With the launch of Heartbeat API, Lana hopes to open a new era of apps for emotion health.

Would you like to be our next guest author? Find out how here!

This is a guest blog and may not represent the views of RE•WORK. As a result some opinions may even go against the views of RE•WORK but are posted in order to encourage debate and well-rounded knowledge sharing, and to allow alternate views to be presented to the RE•WORK community.

Learn more about the future of emotional AI at at the Deep Learning in Healthcare Summit on 25-26 May. The summit will be held alongside the annual Deep Learning Summit in Boston.

Confirmed speakers at the summit include Christhian Potes, Senior Scientist, Philips Research; Sergei Azernikov, Machine Learning Lead, Glidewell Labs; Hossein Estiri, Research Fellow, Harvard Medical School; Mason Victors, Lead Data Scientist, Recursion Pharmaceutical; Muyinatu Bell, Professor, John Hopkins University, and more. Register your place here.