Description
Naive judges rate a few seconds of recorded behaviour on set dimensions, and their ratings are correlated with judgements from long exposure to test whether short excerpts suffice.
The thin-slice method asks how much of a recorded behaviour stream is needed to judge it. In the original study, three 10-second clips were cut from the beginning, middle and end of each of 13 teachers' classes; naive judges rated them on 15 dimensions such as confident, active, warm and competent, and their ratings correlated highly with end-of-semester evaluations by the teachers' own students (confident r = .82, active r = .77). The rigour lies in the criterion: short-excerpt ratings are not trusted until shown to correlate with ratings from long exposure or with outcomes, dimension by dimension. For research operations it offers a way to validate cheaper coding: if raters of 30-second clips agree with full-session coders, analysis can be sampled. It also exposes how quickly first impressions of an agent, voice or presenter form.
Attributes
- Discipline
- Research Design & Rigor
- Loss type
- Unmigrated
- Stage
- Discovery, Evaluation
- Data
- Quantitative
- Effort
- Days
Steps
- Record whole sessions of the behaviour of interest (sessions, calls, demos, agent conversations).
- Cut very short clips at fixed points, for example three 10-second clips from the beginning, middle and end.
- Have naive judges, blind to the criterion, rate each clip on a defined set of dimensions.
- Obtain a criterion from people with long exposure (full-session coders, regular users) or from outcomes.
- Correlate the averaged thin-slice ratings with the criterion, dimension by dimension, and check inter-judge agreement.
- Adopt thin-slice coding only for dimensions that correlate strongly with the criterion.
Sources
- MIT Psychology (Stangor, 2012)
- Physiology of Behavior (Carlson & Birkett, 2017)
Also known as
- Thin-slice judgement validated against a criterion
- Thin slices
- Ambady-Rosenthal thin-slice method
- Brief-excerpt rating
Categories
- Research Operations
- Research Design & Rigor