Article #25
1998
 
 
 
Newsletter list 

Newsletter 

Previous Article 

Next Article 

 
 
 
 

Overconfidence, causal systems, and other people's feelings

Joshua Klayman
Chicago, Illinois

I am working on three loosely Brunswikian projects.

1. A revival of a very old project on overconfidence with Jack Soll and Claudia Gonzalez-Vallejo. We finally figured out (we think) what to make of some data we collected several years ago. Using new analytical techniques, we find that there is indeed some overall bias toward overconfidence, but it varies greatly with what domain you are asking about, how you ask (confidence in a two-choice question or setting a confidence interval), and who you ask (i.e., there are stable individual differences). We do not find any evidence for an effect of difficulty on overconfidence when we control for the kinds of confounds Gigerenzer and colleagues warn about.

2. Continuation of a middle-aged project with Alex Wearing. We have run two studies now on how people learn the causal relations in interconnected causal systems. Using simple three- and four-variable systems, our preliminary findings are that people are pretty good at picking up the correlational structure of the system, but they have trouble distinguishing the possible causal relations underlying the covariations they observe. In particular, people tend to see direct causation where there is only indirect, and causal links where there are only "spurious correlations." With grad student Boris Brodsky, we also have preliminary data on people's ability to think of causal loops as explanations. I think we'll go on to examine what strategies of experimentation and observation are most or least effective for extracting the causal structure from observed covariations in systems.

3. Early work on a new project with various people perhaps including Chip Heath, Chris Hsee, and George Loewenstein. We're thinking about the cues people use in trying to figure out other people's feelings, reactions, preferences, etc. (We think very similar processes may also apply to predicting one's future self and explaining one's past self.) We will try to tie together a variety of findings in cognitive and social psychology by suggesting that there is a consistent pattern of errors that follows from the cues people have available to them, and how they fill in missing or unreliable information.

Contact Joshua Klayman

Previous Article                Next Article

Newsletter list