Methodology and reproducibility
From observation to interpretable evidence
A proposed mixed-methods approach combines carefully scoped qualitative observation with descriptive analysis of voluntary, ethically sourced social interactions. The priority is interpretability, participant welfare and reproducibility rather than collection volume.
Design principles
- Specify the question first: Define the behavioural construct and rival explanations before choosing metrics.
- Collect minimally: Prefer aggregate participation measures and opt-in self-report over granular personal records.
- Separate administration and research: Logs needed to operate the game are not automatically research-authorized datasets.
- Respect context: Account for player expectations, small-group visibility, power dynamics and the operator/researcher dual role.
- Report uncertainty: Retain negative cases, confounders, sparse samples and clear limitations.
Illustrative proposed research workflow
| Stage | Planned activities | Evidence or control |
|---|---|---|
| 1. Design | Formulate research questions; prepare a protocol, ethics determination and recruitment material. | Versioned protocol; approval or documented determination where required. |
| 2. Feasibility | Validate available, non-identifying event summaries and check whether questions can be answered without intrusive data. | Data dictionary; collection minimization assessment. |
| 3. Participant invitation | Recruit adults voluntarily if appropriate; communicate purpose, withdrawal options and use of information. | Consent records stored separately, with secure access. |
| 4. Observation | Collect scoped interaction indicators or researcher notes only within the authorized protocol. | Timestamped field notes and a documented coding guide. |
| 5. Analysis | Compute descriptive summaries and compare alternative explanations; perform thematic analysis where applicable. | Reproducible scripts, interpretation notes and error checks. |
| 6. Dissemination | Publish limitations and aggregate results, avoiding inadvertent identification of individuals or small groups. | Disclosure review and retained methodological notes. |
Operational definitions under consideration
- Social encounter
- A defined event where two or more research-consenting participants are observed interacting under a documented event rule, rather than inferring a relationship from mere co-location.
- Cooperative activity
- Observable assistance or coordinated participation in a shared task; does not automatically imply friendship, altruism or subjective trust.
- Repeated contact
- Subsequent interaction between the same participants within a predefined time window, only when permitted by the ethics/consent protocol.
- Belonging
- A participant-reported sense of connection, not inferred solely from frequency or duration of visits.
- Participation persistence
- An explicitly defined and aggregate-compatible return metric with clear denominators and censoring rules.
Approach to analysis
Quantitative: Proposed analyses include counts, rates, bounded time-to-return summaries, exploratory confidence intervals and robustness checks. Any modelling will be proportionate to sample size, with documented missingness and no post-hoc claims of confirmatory testing.
Qualitative: Field notes or consented interviews may be coded using a documented codebook. Coding changes and contradictory examples will be retained; individual interpretations will not be framed as clinical classifications.
Technical: Analysis scripts may use open statistical tools, tabular processing and reproducible notebooks. AI-assisted drafting or analysis will be reviewed by the investigator; no model-generated findings will be represented as empirical observations.
Data governance and reproducibility
- Separate operational account information from optional research records; avoid collecting private chat, credentials, direct messages or sensitive traits by default.
- Use data minimization and limit collection to a documented purpose. Report small-cell suppression and residual identification risk.
- Record provenance, variable definitions, consent status and transformation steps for any approved research dataset.
- Apply access restrictions, encryption where appropriate, planned retention limits and deletion procedures before participant research begins.
- Share codebooks, aggregate tables and scripts when feasible, rather than sharing identifying raw participant data.
- Keep protocol versions and record whether changes occur before or after seeing research data.