Emotion Intelligence Emotion Recognition Artificial Intelligence

5 Reasons Why Artificial Intelligence Needs Emotion


By: Rana el Kaliouby, Co-Founder and Chief Strategy and Science Officer

I just got back from SXSW where I was on a panel entitled “Can AI Systems Really ‘Think?’” along with John Markoff (of NYT), Adam Cheyer (founder of Siri, now Viv) and Oren Etzioni(at the Allen Institute for AI). The room was fully packed 15 minutes before we started and it was standing room only as people continued to flow in.

I was excited to discuss Artificial Emotional Intelligence, bringing a fresh perspective on Artificial Intelligence, one that very much resonated with the audience. Our point of view is: if AI systems are expected to interact with humans (which many are) then they need to have emotional intelligence (EQ). We also take a very pragmatic approach to AI – we believe that Emotion AI has real commercial applications today and that the benefits to society that come from these use cases far outweigh concerns about a Skynet-like AI taking over humanity.

So why does AI need emotion?


Emotional intelligence is a key component of human intelligence. To effectively reason, plan and perform tasks (all things we think of as cognitive), human beings need to have emotional intelligence. Namely, they need to be able to read and interpret the emotions of others and respond appropriately. In fact, we know that people who have severe problems with EQ are often unable to make key decisions in their lives, keep jobs, or build relationships. On the flip side, people with high EQ are more likeable, persuasive, and have more successful careers and personal lives.

We argue that the same is true for AI: a truly effective AI will need to have emotional intelligence. This is especially true for an AI system that will need to interact, communicate, and coexist with humans, which is increasingly true for AI systems. For example, a social robot or virtual nurse need to have cognitive AI. But without an emotion AI as well, these systems will be unable to truly understand their users, build a strong rapport with them, or be effective in bringing about behavior change.

At Affectiva, we are building that Artificial Emotional Intelligence platform. It has at least two emotion recognition components: sensing and adapting. The AI’s emotion sensing can be done through a variety of sensors (e.g., a webcam), along with prior knowledge about how humans behave and express themselves. It can then adapt its actions, inferences, decisions, and beliefs based on the emotional state of humans, and maybe even other AIs, that it’s interacting with.


There are many esoteric and philosophical issues in AI that we can sit and debate forever. Although this kind of discussion can be fun and stimulating, we are much more interested in a pragmatic view of AI–one that is applied, is very practical, and can benefit society and humanity now. Our point is that there are already very practical use cases of emotion-aware AI, and systems that interact with humans need some level of emotional intelligence. AI systems that are emotion-aware will have an immense competitive advantage over AI that has only cognitive or computational abilities, especially in real world usage.

Social robotics is an area where social and emotional intelligence skills are key. Social robot Tega, developed in the MIT Media Lab as a learning companion for children, is already powered by our emotion recognition SDK. Tega is designed for long-term, in-home and classroom interactions with children, with applications in early-literacy education from vocabulary to storytelling. In the automotive industry, cars are expected to understand the emotional state of drivers – for example drowsy or distracted drivers. Self-driving cars will also need to recognize the state of their human drivers, recognizing if passengers are anxious or worried.


Trust and ethics are a huge consideration in AI – one that we take very seriously at Affectiva. This is especially true when we are talking about people’s emotions – something very personal. If AI systems are expected to make decisions or act on your behalf, what are they allowed to do or not do? Share or not share? If your family robot detects that your teen is showing signs of depression, who should it disclose this to? Building AI systems that have empathy with clear ethical guidelines is critical to the decision-making logic of these AI systems. Emotional intelligence is a key part of that.


Often, the dialogue around AI is one of Human versus Machine! We take an alternative view – one of augmented intelligence. Emotion AI systems will augment human capabilities in a way that extends our capacities, making us more productive and efficient. Take healthcare for example, where there is a shortage of nurses, clinicians and mental health workers. Virtual nurses and home-based robots that care for chronically ill patients won’t replace the role of the nurse or doctor. Rather they will extend it, providing longitudinal and continuous data on the patient’s physical health and emotional well being. Armed with this intelligence, these AI systems can care for patients in a way that is just not possible today. Not only that, it’s scalable, allowing clinical practitioners to efficiently care for a larger group of people.


Robots and other AI systems will take over some of our jobs, but they will undoubtedly create new jobs and opportunities that did not exist before. This has happened over and over in the history of industrialization and automation, and there is no reason to think it will be different in this wave of innovation. I believe we are very quickly approaching a world where AI systems will work alongside humans and extend the reach of what people can do. These AI systems can simply be automation systems that are just software or avatars, or perhaps they will be embodied in robots, devices, or machines. Either way, it will augment what people can do and extend their reach and efficiency. And I think that’s an exciting future.

Emotion Intelligence Emotion Recognition Artificial Intelligence