View source for Bartneck, C. et al. (2009). My Robotic Doppelgänger - A Critical Look at the Uncanny Valley

This paper by C. Bartneck et al. aims to empirically test two predictions made by the “uncanny valley” hypothesis, proposed by M. Mori. These predictions are as follows: highly realistic robots are liked less than real humans (1), and the highly realistic robot’s movement decreases its likeability (2). To put these statements into a more testable form, the authors proposed three hypotheses, which they then proceeded to assess:

1. Androids that are distinguishable from humans will be liked less than humans.

2. A fully moving android will be liked differently compared to an android that is limited in its movements.

3. Androids with different levels of anthropomorphism will be liked differently.

In the paper, three variables are of utmost importance. These variables are likeability, anthropomorphism and the level of anxiety towards robots. The variable likeability is derived from the Japanese word shinwa-kan. Since this word has no exact translation to English, the definition differs between papers. Most papers translate the term to familiarity, Bartneck et al. use likeability. 

Likeability is measured using a scale developed by J. Monathan (Cronbach’s Alpha = .68). Anthropomorphism is measured using a questionnaire developed by Powers and Kiesler (Cronbach’s Alpha = .85). Furthermore, the variable relating to anxiety towards robots is assessed using the robots anxiety scale, or RAS.

To test the hypothesis, a 3 (anthropomorphism) x 2 (movement) experiment was introduced. The three anthropomorphism conditions are masked android/android/human (within-participants) and the two movement conditions are full movement/limited movement (between-participants).

After having performed the experiment, the researchers reached the following conclusions regarding the three respective hypotheses:

1. There is no statistically significant difference in likeability between humans and human-like robots. 

2. There are no significantly different likeability ratings for a fully moving or partially moving robot.

3. There is no conclusive evidence that androids with different anthropomorphism levels are liked differently.