Wall-E is cute, and Baymax is endearing. The Stop & Shop robot is annoying, but googly-eyed and harmless. The vaguely humanoid robot that seems to be looking at you, on the other hand, is terrifying. This is what’s known as the “uncanny valley effect,” a hypothesis proposed in the 1970s by Masahiro Mori, a Japanese professor of robotics. The theory states that as robots possess more humanoid traits, people tend to feel more empathetic and positive towards them. However, as soon as a robot looks almost fully realistic with subtle flaws, people's responses will shift abruptly into revulsion, discomfort, or fear. But why does this phenomenon exist in the first place, and what impacts could it have on the greater scientific world? A number of hypotheses exist, and while no definitive conclusions have been drawn, a team of researchers led by Jari Kätsyri at Aalto University in Finland sought to summarize the current state of the field as it is known.

After a comprehensive literature review, the team found that the perceptual mismatch hypothesis is the only existing hypothesis strongly supported by present literature. This theory, coined in 2009 by researcher Karl MacDorman, states that the unease associated with the uncanny valley is due to inconsistencies in the realism of anatomical features. A clearly artificial mouth on an otherwise human face would be an example of the perpetual mismatch hypothesis. In contrast to this, the study was unable to find strong evidence for the categorization difficulty hypothesis, which states that the uncanny valley is experienced when we are unable to subconsciously group an object into a clear and familiar category, and for which movement is a strong effect modulator.

The uncanny valley effect may have a greater evolutionary explanation as well. A 2010 paper from the University of Bangladesh suggests that, because humanoid anomalies to some extent resemble corpses, we as a species have evolved to be wary of such features, since avoiding dead bodies has historically protected human populations from pathogens and disease.

More recently, this phenomenon has proven to have unintended effects on the world of scientific research. A team of researchers led by Anna Flavia Di Natale at the University of Milan looked into how virtual faces that have replaced photographs as stimuli in human face perception studies could be skewing results. The team found that many virtual faces used in these studies do in fact elicit the uncanny valley effect, and that researchers should be more cautious when using computerized alternatives in studies involving the human face. Notably, these concerns don’t just end in the lab — with the escalation of AI implementation into daily life and increasing human-automaton interactions, the uncanny valley effect grows ever more relevant in our world today.