Understanding the Empathy of Artificial Intelligence
Samantha Carollo is a participant in the OREL undergraduate research intern program. This research is part of OREL’s Society, Ethics, and Technology theme.
Empathy is the ability to understand and share another person’s feelings. While we do not need to intentionally think about being empathetic before doing so, it is nevertheless a core component of our everyday behavior. Often, exhibiting empathy is often as automatic as breathing. For Artificial Intelligence (AI) to become more human-like, it must master empathy.
Humans utilize four types of intelligence: mechanical, analytical, intuitive, and empathetic. All these types of intelligence are required for AI to behave something resembling human (Huang & Rust, 2018). While mechanical, analytical, and intuitive intelligence are very easy for AI to replicate, empathetic intelligence is not. This is because AI training generally occurs in the form of repetition and reasoning. By contrast, empathy is emotional, and cannot be acquired through training. Pure, raw emotion is what allows humans to connect with other humans in a deep and meaningful way that a machine cannot yet understand.
Of the four types of intelligence, empathetic intelligence is the most difficult for AI to replicate. As hard as it is for humans to process each other’s emotions, this task is much harder to implement in robots. The task of creating empathetic AI involves seems simple enough: describe a machine that can generate emotional responses. Yet to be successful at generating the correct emotional response, empathetic AI also needs the ability to analyze and understand human emotions. Empathy is the most important thing for Artificial Intelligence to have because emotions are a critical component of how humans process everyday situations. In fact, if an AI cannot process empathy, then it will simply be unsuccessful in everyday life.
Three key components of empathy need to be replicated in order for AI to replicate the everyday behavior of humans: emotional empathy, motivational empathy, and cognitive empathy. Emotional and motivational empathy help humans both experience the world around them and figure out how to cope with everyday challenges. Cognitive empathy is used to help others recognize their emotions and process the situation they are in, and ultimately in determining how to proceed (Montemayor et al. 2021). These three components make up the multidimensional nature of empathy, without which human-level intelligence would seem quite different.
“Empathic behaviors are thought to be learned through social interactions with humans in the framework of (cognitive) developmental robotics” (Asada, 2015).
Another reason to discuss the concept of empathetic intelligence in AI has to do with job automation. An ever-increasing number of jobs are becoming automated. As job automation will affect many aspects of human life and livelihood, it will also change the workforce as we know it. Think about customer service/retail jobs. The self-checkout line might be more convenient, but results in one less human is being paid to do this job. In the field of healthcare, where empathy is a key skill, people are going to feel less comfortable in tragic or crisis situations.
The main need for empathetic AI in an automated workforce is to avoid the serious risk to humans when an AI cannot handle the emotional demands of a particular job. As Montemayor et.al (2021, pg. 5) put it: “autonomy and agency are the hallmarks of attentive agents — this is how their mental actions are not simply accidentally correct — and only attentive agents can care for others”. In a positive development, some groups that develop healthcare AI systems are being cautious and incorporating some form of empathetic capacity into their applications.
Emotions are what make human life exciting, from a child laughing in the playground to an adult having a passive-aggressive argument at work, emotions are what make life and without the empathy to understand those emotions life seems empty. They are also essential from a broader cognitive standpoint as well: empathy is intimately related to emotional processing in the context of understanding child’s laughter as happiness or an argument at work as passive aggressive. While there’s no predicting whether future artificial intelligence will be able to generate and process empathy at human levels, we must consider that today’s technology is lacking. The ability of AIs to process human emotion is the final step in making them more human-like, and this means opening the next chapter in this emerging technological field. Once empathetic intelligence becomes an unquestionable part of AI applications, AI itself becomes less a concept of science fiction and more a vital part of our reality.
Further reading: Artificial Empathy (Wikipedia), Uncanny Valley (Wikipedia), Robot Ethics (Wikipedia)
References
Asada, M. (2015). Development of artificial empathy. Neuroscience Research, 90, 41–50.
Davenport, T., & Kalakota, R. (2019). The potential for artificial intelligence in healthcare. Future Healthcare Journal, 6(2), 94.
Huang, M.-H., & Rust, R. T. (2018). Artificial intelligence in service. Journal of Service Research, 21(2), 155–172.
Montemayor, C., Halpern, J., & Fairweather, A. (2021). In principle obstacles for empathic AI: Why we can’t replace human empathy in healthcare. AI & Society, May.
