Emotional Machines

Emotions are one of the most unique qualities that we have as a species. They are a constructed social reality based on our brain’s sensory understanding of the world that has allowed us to grow as a society to the point where we are today (Something that no other known species has ever done). Emotions are a key element for us to communicate, cooperate and grow, they are the basic ingredient of our outstanding social nature. Without emotions, we would not be able to nurture our believes, our values nor our culture. In other words, society would not exist as it currently does. Emotions have been a true pillar for human endeavors throughout our entire history and they must be taken into account within the upcoming exponential advancements of technology. 

It is clear to me that if we want to truly humanize technology and create a new type of symbiosis with it, one that manages to benefit the most humane endeavors of our lives, it must be able to understand and deal with human emotions. Our systems must actively be aware of how we are feeling so that they can respond to meet our needs in the most appropriate and humane way possible. Emotions are key to human understanding and they are the base of our interactions, therefore we must create human-machine interactions that have the same emotional context and meaning as human-human interactions.

Imagine the benefits that robotic hospital rooms, offices, homes, streets and parks could have on our lives if they were controlled by intelligent systems capable of understanding how users feel and proactively change their physical state to a state that amplifies or diminishes their user’s current emotion depending on their specific needs. Patient recovery would be faster, work would be more enjoyable and productive, stress levels would be reduced, streets would become more walkable and human quality of life would rise overall. 

Now imagine autistic children interacting and learning with robot tutors that are able to create more pleasurable learning experiences through the navigation of the child’s emotional states or maybe creating robots that are able to empathize with mentally ill patients or that serve as care givers to elderly people, machines that are able to grasp emotional context would help them stay in better health and have a more enjoyable life. All of this ideas sound compelling but to make them happen more research has to be done to fully develop the described technologies. 

Recent discoveries in neuroscience have given clear evidence that emotions are the result of a construction process based on error and prediction carried out continuously by the brain throughout our lives, this means that emotions are created by our brain’s understanding of multiple sensory inputs that describe our diverse physical and mental states. Our physical states are mostly created by our body’s metrics (temperature, blood pressure, perspiration, cardiac rhythms, brain activity, etc), the environment that surrounds us (temperature, humidity, light exposure, pressure, movement, etc), and the context in which we are in (current interactions with others as well as events happening around us) the mental states are the interpretations that we make of the physical state (previous experiences and memories).

Lisa Feldman gives compelling evidence in her book “How Emotions Are Made: The Secret Life of the Brain” arguing that emotions are not natural responses or instincts that our body has from the first day of our lives but rather a set of instances that we learn to classify as we grow and interact with society and culture. This means that emotions are the result of years of cumulative learning carried out by humans! The described evidence constitutes the basic assumption for the theory of constructed emotion described in the book, a theory that has sparked several ideas in my mind about ways in which we can build systems that are capable of understanding human emotions. 

Looking at the theory of constructed emotions from a technologist’s lens, I believe that if the right sensors (environmental, contextual, mental and biological) are developed, we could start to create data sets that resemble the data inputs that our brains use within the process of creating and understanding emotional states. With the right sensors and selection of variables, we could mimic the natural inputs used by our brains to evaluate emotional states (physical state variables and mental state variables). These data sets could then be fed into learning algorithms that will generalize patterns and labels in the same manner that our brains do. 

I believe that the development of emotional systems is in no way limited by the current state of the art of learning algorithms, processing power nor by our current understanding of emotions but by the lack of emotion data labeling techniques and sensors that will help us integrate appropriate data sets (obtained through clever and informed processes) for learning algorithms to generalize with.

The development of this systems is not too far of into the future and therefore it is imperative that we start concerning ourselves with the negative impacts or uses that this technologies could have. The first (and in my opinion the most relevant) misuse of the technology that comes to my mind is mass manipulation of society through emotions (disturbances in consumer behavior, politics, education, healthcare and business are evident and horrifying). As stated above, emotions are a crucial component of our behavior and our social reality, they are the most powerful influencers of our actions. If systems are capable of understanding our emotions, they might also easily become capable of manipulating them so that our actions benefit the people in control of the systems (companies, governments, etc).

It is my belief that to avoid dystopian scenarios of emotional manipulation the development of emotional systems must be done with the ideals of open source communities; data sets, source code, experiments, applications and real life implementations must all be shared and accessible to anyone who wants to use them or inspect them, in this way systems do not have centralized control and people from all around the world can keep the systems free of manipulation (a similar principal to the one that makes blockchain so attractive). In addition to an open source based development, there needs to be an integration of people with truly diverse professional and cultural backgrounds during the entire development process. The addition of a diverse set of skills and cultures will keep the systems free of bias, equitable and democratized. 

Integrating emotions to exponential technologies will be key to maintaining a balanced relationship between humans and technology. I foresee that in the future, systems that are able to adequately understand emotions will be the systems with the highest probability of creating a positive impact in our every day lives due to the power that emotions have shown throughout centuries. The proper development of our most humane qualities through a symbiosis with technology relies on a flawless design of the emotion based interactions that we have with other humans as well as with the machines that will be present in our everyday lives.  

Andres Rico

Invention that embodies the ideas described in the text: emotIOn.