Measuring prosocial behavior
A measure is a sample of behavior or experience under particular conditions. Choose it for the claim you want to test. A questionnaire about concern, an intention to volunteer, and completed volunteer work are different outcomes.
This primer is a methods guide for the SAGE Lab. Use it alongside the field overview and altruism primer.
Specify the target before the instrument
Record the intended beneficiary, the actor’s available options and resources, the opportunity to act, actual costs, and the outcome of interest. Decide whether the claim concerns a state, a decision, a repeated pattern, an intervention effect, or an underlying motive.
Example: “participants allocated more of their bonus” is a narrower claim than “participants became more altruistic.” Evidence for the latter needs a definition of altruism and a design that addresses alternative motives.
Common approaches
| Method | What it samples | What needs checking |
|---|---|---|
| Dictator allocations | Unilateral sharing of a defined resource | Stakes, ownership, recipient, anonymity, framing, and understanding. |
| Trust and public-goods games | Decisions involving dependence, possible returns, or collective outcomes | Beliefs, strategic incentives, risk, and the specific interaction structure. |
| Actual helping tasks | Assistance, effort, time, or task substitution | Whether declining is genuinely possible, help is needed, and the action has consequences. |
| Giving and volunteering records | Behavior with practical consequences | Resources, exposure to requests, access, and unequal opportunities to contribute. |
| Questionnaires and informant reports | Recalled behavior, self-perceptions, motives, or observed tendencies | Item content, timeframe, reporting bias, and who could observe the behavior. |
| Experience sampling | Repeated recent experiences in everyday contexts | Missed opportunities, selective reporting, burden, missingness, and dependence between observations. |
A transfer in a trust game can reflect more than trust alone. Cox (2004) used comparison games to distinguish trust and reciprocity from other-regarding allocation preferences. A consequential choice can still have an ambiguous psychological interpretation.
Reliability and generalization
Repeated trials can improve precision, but repeated versions of one task do not establish transfer to different forms of help. Aggregate across meaningfully chosen occasions, model learning and order effects, and retain information about context.
Evidence about external validity is mixed. Galizzi and Navarro-Martinez (2019) reported limited prediction of field behavior from social-preference games. Wang and Navarro-Martinez (2023) showed that aggregating games and field behaviors can improve prediction. These findings motivate checking measurement error and the target domain rather than treating all games as either universally valid or useless.
Experience sampling brings assessment into daily life. Hofmann et al. (2014) used repeated reports to study everyday moral experience. Such reports remain reports: they do not independently reveal every missed need or establish recipient benefit.
Separate amount from quality
Measure what the actor does and what the recipient receives. More money, time, or effort may be relevant, but benefit also depends on suitability and consequences. Record burden and constraints when interpreting cost.
For intention–action research, distinguish no opportunity, inability to act, refusal, and missing information. If training changes which opportunities people notice or enter, comparing only noticed opportunities may compare different sets of situations.
A practical design checklist
- Define the construct and the exact claim.
- Choose a task with consequences and demands relevant to that claim.
- Measure opportunity and resources separately from willingness.
- Check comprehension, credible incentives, and possible floor or ceiling effects.
- Specify a primary outcome and relevant contrasts before inspecting results.
- Use multiple occasions or methods when making a broad claim.
- Report what was observed and what remains an inference.
For an intervention, an immediate task tests an immediate outcome. Persistence requires later assessment; transfer requires a different task, context, or recipient. A new scale also needs evidence about structure, reliability, and validity; a label and internally consistent items are insufficient.
Essential reading
- Cox (2004): experimental contrasts that clarify the interpretation of trust and reciprocity.
- Galizzi and Navarro-Martinez (2019), with Wang and Navarro-Martinez (2023): contrasting evidence about lab–field prediction and aggregation.
- Hofmann et al. (2014): an example of repeated assessment outside the laboratory.