Carnegie Mellon University HCII Work on these projects has finished and the  last update for this site was in 2010. The site will be maintained as a community resource.
anthropomorphism.org

Psychology: Measuring Anthropomorphism

How do we know when someone is anthropomorphizing?

Here are some of the ways we can measure anthropomorphism. Most of these measures have some problems--such as confounding animism and anthropomorphism, or confounding anthropomorphism and attachment. Thus, measurement is a research question in its own right.

  • Rating scales
  • Narratives
  • Word counts
  • Social interaction and task behavior

 

Rating scales

The simplest way to measure anthropomorphism is to ask people to rate how humanlike an object or animal is. For example:
How humanlike is this robot?

We also have used a human trait rating scale, based on the Big Five personality scales, and a machine trait rating scale [PDF 112K].

A big problem in using rating scales is that people interpret the ratings they are supposed to make differently depending on what or whom they are rating. For example, in rating the cheerfulness of a humanlike robot that told jokes, we think many people compared the robot to other robots, not to people. This tendency meant that people ended up rating a robot as more cheerful than people. We don't really know if they thought the robot was unusually cheerful for a robot, or if they thought the robot was cheerful in the universe of all cheerful beings.

 

Narratives

A very interesting way to evaluate anthropomorphism is to ask people to describe an animal or object's behavior, or to explain it. We have used Malle's coding scheme [PDF 23K] to examine how people interpret the behavior of objects, robots, and people.

Narratives also can be taken from interviews and ethnographic studies. For instance:

Siino, R. & Hinds, P. (2004). Making sense of new technology as a lead-in to structuring: The case of an autonomous mobile robot. Best Paper Proceedings of the Academy of Management, New Orleans, LA, August 2004. Contact Pamela Hinds for a copy. for a copy.

 

Word counts

Word counts and word coding schemes can be used to test hypotheses. For instance, count the number of words that people use in explaining a concept to a robot. The more words they use, the less they think the robot already understands. (Update: Adam Kramer's "Text Analysis and Word Counts" is no longer available on the andrew web system.)

 

Social interaction and task behavior

A powerful form of measurement is to watch and measure a person's behavior as they interact with a nonhuman (such as a robot). An example of this use is found in Experiments 3 and 4 of this paper on personality and task matching.

Goetz, J., Kiesler, S., & Powers, A. (2003). Matching robot appearance and behavior to tasks to improve human-robot cooperation. Robot and Human Interactive Communication, 2003. Proceedings. ROMAN 2003 (pp. 55-60). The 12th IEEE International Workshop on, Vol., IXX., Oct. 31-Nov. 2, Milbrae, CA. [PDF 748K]

See also:
Hinds, P., Roberts, T., & Jones, H. (2004) Whose job is it anyway? A study of human-robot interaction in a collaborative task. Human Computer Interaction, 19, 151-181. [PDF 220K]

 


This website promotes understanding of anthropomorphism. Our work on these projects has finished, but the website will continue to be available as a community resource.