Confederate effect

Term used to describe the phenomenon of a human considered a machine from their textual discourse (See: The Confederate Effect in Human Machine Textual Interaction: Shah & Henry, 2005) [1].

It is the reverse of the Eliza Effect [2] : “our more general tendency to treat responsive computer programmes as more intelligent than they really are” and the cause to “very small amounts of interactivity” resulting in humans to “project own complexity onto the undeserving object” (Shelley Turkel, in Life on the Screen –Identity in the age of the Internet, p. 101, 1997).

In the first Loebner Prize for Artificial Intelligence [3], in 1991, which deployed restricted conversational one-to-one Turing imitation games [4] - each interrogator chatting to one Artificial Conversational Entity (ACE) at a time, a female ‘confederate’ or hidden-human, chatting on the topic of Shakespeare was considered too knowledgeable, hence considered a machine . The phenomenon was seen in the University of Surrey held 2003 Loebner Prize for Artificial Intelligence, when both hidden-humans, one male and one female, were each ranked as machine by at least one judge: Judge 7 and Judge 9 ranked the female ‘confederate 2’ as “1.00=definitely a machine”; the male ‘confederate 1’ was ranked “1.00=definitely a machine” by Judge 4 and Judge 9. [5] The gender of these two hidden-humans were incorrectly identified (male considered female; woman considered man) in independent transcript analysis ('gender-blurring' phenomenon, see Shah & Henry, 2005). [6]

Of the many challenges facing more effective human-machine interface design, in domains that vary from large-scale process-control computing systems, nuclear energy or security uses to Internet pages, ignoring the ultimate user, a human with all their inherent habits and peculiarities, will result in producing little used or at worst failed systems. The need then, to infuse intelligence into interfaces is not only an option but safety may depend on it. We propose a move to design of cognitive interfaces.

A cognitive interface is one that would be able to perceive the gender, age and expertise level of a user, and thus adapt its behaviour to suit them. A cognitive interface would enhance user experience, ultimately leading to more efficacious use of time and tools in human-machine interaction. In order to design interfaces capable of intelligent discernment and ability to balance system’s users, three features present in human perception, when interacting with machines, ought now to be fully considered. They are the Eliza, Confederate and Gender-blurring effects

References

1. * "The Quest for the Thinking Computer".(In: Epstein, Roberts, & Beber) Parsing the Turing Test: Philosophical and Methodological Issues in the Quest for the Thinking Computer, 2008, pp. 3–12