Tuesday, December 13, 2005

The problem with expert testimony

Experiment:

Witness A: "I'm not sure it was the defendant I saw pouring toxic in the lake, but I suppose if I had to guess I'd say it was. "

[Exit witness, who backstage puts on a lab coat and glasses, then returns to the witness box]

Witness B: "There's a statistically significantly greater probability that the defendant did it than that she did not. I'd stake my reputation as Dean of Harvard on those numbers."

Observation: B sounds more persuasive than A

Diagnosis: We conflate statistical confidence with personal confidence. When someone knows exactly how unsure they are, we see them as confident. Note though that the Dean didn't wager so much as a tuna sandwich on the defendant being guilty.

2 comments:

t.s. said...

Is that a problem with expert testimony, or witnesses who know how to say stuff? Granted, the former is usually a subset of the latter.

toggi greek forum said...

Greaat reading your blog post