Description
Designs study tasks by sampling the real environmental conditions and the natural correlations between cues, not just the participants, so findings generalise to real use.
Brunswik argued that psychological processes adapt behaviour to environments, so experiments must sample environmental conditions as carefully as they sample participants. Representative design is often confused with ecological validity. It requires the conditions of a study task to match the conditions in which the behaviour normally works, including the natural intercorrelations among cues (Brunswik's 'ecological intercorrelation'). If a lab task breaks those links, for example by removing an information source that normally co-occurs with another, people may use different strategies than they do in real life. Davids and colleagues use this to argue that the lab-versus-field split matters less than whether key organismic, task and environmental constraints are kept. The Handbook gives examples in which small changes to task constraints, such as batting against a bowling machine instead of a real bowler, changed batters' timing and coordination. For UX the principle turns into a design audit for usability tests and experiments: sample realistic data, devices, interruptions and time pressures, keep the cues users normally combine, and check that the response asked for matches what users really do.
Attributes
- Discipline
- Research Design & Rigor
- Loss type
- Latent
- Stage
- Definition, Evaluation
- Data
- Analytical / expert
- Effort
- Days
Steps
- Describe the target use situation: the information sources users combine, typical interruptions, devices, data and time pressures.
- List the cues and constraints in the planned study task and mark which have been removed, simplified or decoupled.
- Sample a range of realistic situations (data sets, scenarios, contexts) rather than one idealised task.
- Keep natural correlations between cues, for example realistic messy data rather than clean demo data.
- Match the response mode to real behaviour (acting, not only saying what one would do).
- Where full representativeness is not possible, document the departures and treat them as threats to generalisation.
Sources
- Tenenbaum, G. & Eklund, R. (eds), Handbook of Sport Psychology, 3rd ed. (2007)
Also known as
- Representative design
- Brunswikian representative design
- Constraints-based task design
- Task representativeness
Categories
- Research Operations
- Research Design & Rigor