Let’s play a game. Imagine you wake up (or better, your conscience does) and you are trapped in an empty room inside your brain. You can see, move, and touch with your own body, but you cannot speak. In fact, your mouth opens, words flow out; you can laugh and interact with your family and friends, but you cannot decide what to say. You cannot choose your reactions, behavior, gestures, and another figure which you don’t see makes all these decisions. You are yourself in everything except you do not control the way in which you interact with other humans. Would they detect the intruder that controls your social interactions? Would they know that you are not really you?
You would be, in the poetic words of Stanley Milgram, a cyranoid: a human that is controlled by an agent (human or artificial) that interacts with other humans. A cyranoid is a body that pronounces a speech, alien with respect to the source of that speech.
Milgram’s cyranoid and the concept of cyranic illusion are the last chapter of his seminal career in psychology and social science. He crafted an experiment in which various sources (including himself) gave voice to a plethora of bodies, ranging from adult males, to women, to children. In one of the most meaningful rehearsals of this experiment, Milgram himself served as the source to an eleven-year-old kid, the shadower. After conversating with the kid, a group of teachers was not able to spot the oddity in its speech and described a brilliant eleven-year-old with great speaking abilities. This is known as a cyranic illusion: what appears to be a particularly charming kid is instead hosting the thoughts of a veteran of social interactions. The term cyranoid and the adjective cyranic derive from the play character Cyrano de Bergerac, a man that, while being very intelligent and charismatic, believes his physical appearance will prevent his distant cousin from falling in love with him. To solve this problem, he teams up with a handsome but dull man that will serve as the shell for Cyrano’s thoughts and words.
Since Milgram did not publish any paper on the cyranic illusion but only some anecdotal findings, Corti and Gillespie, two social psychologists from the London School of Economics, replicated his experiments with cyranoids in a controlled lab setting. However, they went a step further. They were interested in finding out which agent, when interacting with a human counterpart, could generate the highest degree of intersubjectivity. The term has been defined by the two researchers in various ways, but it generally indicates the intersection between two counterparties’ cognitive perspectives. In other words, it measures how much the human involved in an interaction perceives its counterpart to be of the same nature of himself.
First, Corti and Gillespie found out that when confronted with a cyranoid à-la Milgram, most of the participants in the experiment could not break the cyranic illusion despite the incongruities between the words they heard and the subject that was pronouncing them, thus reinforcing Milgram’s findings. Moreover, Corti and Gillespie also set up one source-shadower combination in which the source was an artificial intelligence, and the shadower was a human. They confronted this agent, which they called echoborg, with the case of an AI that voiced a robot, and they analyzed whether the human counterpart was more willing to initiate verbal repairs to misunderstandings with the first or the second agent. They discovered that when presented with a human body, holding the source of the speech fixed, human counterparts were much more likely to attribute to the agent human-like features, and to engage in verbal repairs with them, therefore demonstrating a higher level of intersubjectivity. Corti and Gillespie explain that the echoborg method should be used to test agent-human intersubjectivity against a human-human benchmark. In other words, to evaluate the human semblance of an AI’s speech and behavior in an interaction with a human counterpart, a human body should be used as the medium through which the AI expresses himself.
Isn’t it paradoxical that creating an AI that perfectly mimics speech, feelings and behavior of humans still results in a weaker human recognition from its conversational counterpart that using a chatbot (as Corti and Gillespie did) to voice a human body? The applications of cyranoids could space multiple disciplines and are generally oriented towards examining how humans engage in communication patterns when confronted with different counterparts, and how these patterns change when the source of the interaction remains fixed but the physical body that participates in the interaction changes. This can highlight all the cognitive biases that make human communication suboptimal in a multitude of contexts. The most fascinating applications are, however, those of the recognition and the humanization of artificial intelligence.
Corti, K., Gillespie, A. The Body That Speaks: Recombining Bodies and Speech Sources in Unscripted Face-to-Face Communication. Front. Psychol., 08 September 2016.
“Cyranoids”: Stanley Milgram’s Creepiest Experiment, Discover Magazine, 06 September 2014.