Highly Humanlike Machines & Uncanny Valley

Ernst Jensch

• intellectual uncertainty

• not being at home in the concerned situation

• are living beings really alive and vice versa are lifeless objects really not alive

  –> tree appearing like a giant snake when it is moving

Sigmund Freud

• unfamiliar but even though familiar at the same time

• remembering back to the bad experiences in our childhood

• the phenomenon of the Doppelgänger

  –> modern Doppelgänger, very similiar looking robots

Masahiro Mori

• There is an area between highly human likeness and no simliraty to humans, where robots are exceptionally uncanny, creepy

• This is also shown partly in research later on

The uncanny valley hypthesis pictures a graph in which the familiarity to robots increases as their similarity to humans increases until a certain point from which on slight differences and imperfections suddenly are perceived as creepy and eerie

But complete human likeness is of course perceived as familiar and not dangorous

They are not perceived as uncanny

Range from abstract/mechanoid machines to slightly humanlike

Well distinguishable from real persons

Sometimes cute/Kindchenschema designs, especially in social „companion“ robots

An example would be the personal robot “RoboHon”

Not perceived as uncanny, because not distinguishable from real humans anymore

Ethical question: Do we want to live in a world where humans and machines are impossible to distinguish?

–> EU AI Ethics Guidelines say: Machines must be identifiable as such

Examples:

• Thispersondoesnotexist.com

• lifelike synthetic voice

Perceived as eerie, spooky or even threatening

Examples:

• Animationmovies like Polarexpress, since particularely the eyes are not portrayed realisticaly

• Humanoid robots like Telenoid or Ameca

Evolutionary explanations

Non-perfect android figures remind us of disease or they remind us of our own mortality

Cognition-based explanations

• Categorical Uncertainty

  Not clear to which category the figure belongs (Human or machine? Alive or inanimate?), ambiguity about the presence of threat (Uncanny feeling enables us to stay alerted until clear)

• Prediction error

  If people evaluate the robot’s behavior according to a human schema, it might not measure up to these expectations due to ist imperfections

It is the phenomenon that machines can also occur uncanny even if there is no visual image of a robot

Machines described as having the ability to make their own decisions and actions (“agents”) are perceived as more uncanny then machines described as being controlled by humans (“tools”)

Machines described as being able to feel some form of emotions, having sentience and personality (“experiencers”) are perceived as even more uncanny

It is shown that AI is mostly portrayed as either some humanoid robot or as an abstract graphic

Thereby humanlike pictures of AI strongly correlate with the images which are perceived as most thratining or creepy

But most of the images or graphics which show up, mostly do not really have to do something with usecases of AI and are therfore not considered as an realistic portrayal of the topic

 People report to be unfamiliar with “AI” or “machine learning”

Numerous studies show that little knowledge and the feeling of lacking competence are major drivers behind tech anxiety

People who are afraid of something or feel that AI is not for them will not deal with it constructively, but avoid confrontation

Current media narratives and media images either raise unrealistic expectations (AI as a magic solution to almost all problems) or they fuel fears (conscious machines that will soon outperform humans in everything)

AI Literacy begins with a basic understanding of what AI is, the language surrounding the technological and social aspects of AI, how AI works and how it is currently playing a role in our daily lives, in addition to the potentiality of future implementations in all professions and industries

Because people need to be empowered to participate in the discourse about our technological future

Because people should be able to understand when they may be affected by algorithmic decisions

Because people should get a sense of how they can use AI in their own field of action and what benefits could be derived from it

Because people shall have a more realistic view of opportunities and risks related to AI