All of us have heard this several times: robots and artificial intelligence (AI) will replace many of us, and this also goes for service employees who used to have face-to-face interaction with customers. Much research has already been carried out on what the new interfaces should be like to make us humans feel comfortable in our roles as customers. So far, one result is particularly dominant: the more the interface can be made to resemble a human, the more comfortable we feel.
Such results stem from studies in which researchers have measured customers’ perceptions of humanness of robots and other agents as well as customers’ evaluations of them. And the typical result, “what-is-humanlike-is-good”, is one main reason why most AI-powered virtual agents are designed so that they have a humanlike face, a name and a gender, and that they can talk to us in a conversation mode. In other words, human attributes have become a gold standard for what the new, virtual entities should be like.
If real humans are assessed in the same way as robots and virtual agents, in terms of perceived humanness, one may perhaps think that humans would receive “full” humanness scores. A human is a human, right? However, after having seen the results of copious studies in which participants were asked about how they perceive various humans in terms of humanness, and after having made some studies on this issue myself, I have so far not found one single study in which real humans received full humanness scores. In other words, perceived humanness is a variable, not a dichotomy, and it rarely reaches its highest possible value when people are asked about the humanness of other people.
Perceived humanness is a variable, not a dichotomy, and it rarely reaches its highest possible value when people are asked about the humanness of other people.
The main reason is probably that being human means having many characteristics related to an inner life – such as having agency, an ability to experience emotions, and having a feeling for what is right and wrong – and these characteristics will never be fully accessible to others. So, no one can know for sure what it is to be me, and I cannot fully know what it is to be someone else. This is indeed an essential part of being human – and a timeless source of pleasant surprises and disappointments.
In any event, studies showing that real humans are typically denied full humanness by others are often made under the label of dehumanization. This is a malicious form of perceptual bias because it can lead to many horrible outcomes – such as genocide, torture, slavery, rape, and various forms of other anti-social behaviors. The main idea is something like this: if another person is not a full human, well, then we have a green light to do whatever we want with this person.
Lately, I have had a chance to see what happens if you move beyond those groups that are studied in the dehumanization literature (homeless, migrants, drug addicts etc.). More specifically, my idea was to examine customers’ interactions with service employees through the lens of perceived humanness, and I wanted to know what would happen when customers are explicitly asked to assess service employees in term of perceived humanness.
So, for example, in one experiment I was inspired by the current corona pandemic and the new norms for behavior it has generated. The design was such that the participants were encountering a female grocery store employee who existed in two versions. In one version, she violated the norms for what is appropriate social behavior during a pandemic by coughing into the air and by neglecting social distancing; in the other version, she followed the norms. With this design, then, the participants were randomly allocated to encountering one of the two employee versions, and after the encounter I asked questions about their reactions.
Not so surprising, the norm-violating employee was dehumanized – in the same way as migrants, garbage collectors, and homeless people have been in other studies. This part of the results should be seen in the light of people’s general sensitivity to whatever is signaling disease; if there is a suspicion that someone is infected, we typically react with avoidance and disgust (and dehumanization). More surprising, however, is what happened to the participants who were exposed to the norm-complying employee – because this employee was dehumanized, too. Presumably, one reason is that decades of exposure to the idea that “the customer is king” can devaluate those who have a service role.
The take-away, I believe, is this: So much attention has been devoted to making non-human entities more humanlike that we may have forgotten that we humans typically are subject to a humanness deficit in the eyes of others. And this is far from optimal when a “what-is-humanlike-is-good” mechanism is operating. The service firm in which human employees will still interact with (human) customers, then, should be mindful of what is boosting the perceived humanness of human employees.
Professor of Marketing and Head of the Center for Consumer Marketing (CCM), Stockholm School of Economics
Senior Fellow Researcher at CERS, Hanken