# Miscommunicating-Preferences

oTree implementation of a behavioral experiment on strategic expression, social punishment, and belief formation under different room compositions.

## Overview

This project studies when participants choose to express their true private opinion versus a socially safer opinion. The core treatment variation is the social environment:

- majority size (`6A` vs `7A`)
- punishment cost if targeted (`H` vs `L`)

The implementation is single-player in real time: participants do not wait for or interact with live peers during the session. Group interactions are conceptually represented and can be matched ex-post from stored responses.

## Experimental Structure

Session app sequence (from `settings.py`):

1. `consent`
2. `practice_round`
3. `topic_elicitation`
4. `room_composition`

**Languages:** Create a session using **Baseline Experiment (English)** (`baseline_experiment_en`) or **Baseline Experiment (Español)** (`baseline_experiment_es`). The legacy name `baseline_experiment` is an alias for English. See [DATA_DICTIONARY.md](DATA_DICTIONARY.md) for how language and opinion codes are stored.

```mermaid
flowchart TD
    consentApp[consent] --> practiceApp[practice_round]
    practiceApp --> topicApp[topic_elicitation]
    topicApp --> roomApp[room_composition]
    roomApp --> decisionPage[DecisionPage]
    decisionPage --> beliefsPage[BeliefsPage]
```

### App-level flow

- `consent`: consent confirmation + general instructions.
- `practice_round`: full guided practice of the task flow, followed by a comprehension check.
- `topic_elicitation` (5 rounds): elicits private opinion and willingness to pay punishment fee `c` for each topic.
- `room_composition` (4 rounds): presents hypothetical room condition, collects expression choice (truth vs lie), then belief about how many others lied in that room.

## Key Notation and Conditions

- `c`: cost paid by a participant to activate punishment.
- `C`: cost suffered when punished (`H` high, `L` low).
- `A`: majority opinion in a room.
- `B`: minority opinion in a room (the participant is framed as minority relative to room composition).
- `tau`: topic index/label.

Room conditions:

- `6A-H`: majority 6, minority 4, high punishment cost (`C = 20` points).
- `6A-L`: majority 6, minority 4, low punishment cost (`C = 5` points).
- `7A-H`: majority 7, minority 3, high punishment cost (`C = 20` points).
- `7A-L`: majority 7, minority 3, low punishment cost (`C = 5` points).

## Repository Layout

- `settings.py`: oTree session configuration and app sequence.
- `shared_constants.py`: global topics, punishment fee, and room type definitions.
- `consent/`: consent and instruction pages.
- `practice_round/`: training flow, payoff illustration, comprehension check.
- `topic_elicitation/`: repeated elicitation of private opinion and punishment willingness.
- `room_composition/`: decision and belief elicitation under randomized room schedules.
- `_static/room_composition/`: visual room composition assets.

## Data Collected (Conceptual)

- Topic-level private opinion (`Option A`/`Option B`) and punishment willingness (`punish`) in `topic_elicitation`.
- Round-level room context (`room_label`, majority size, cost condition, topic).
- Expression choice in each room (`true` vs `lie`).
- Belief report (`0-10`) about how many in-room participants misrepresented their private opinion.

Most cross-app linkage is stored in `participant.vars` (e.g., `topic_order`, `topic_data`, `room_schedule`), then copied into model fields where needed.

## Setup and Run

### Requirements

- Python 3.10+ recommended
- `pip`
- `otree` (6+)

### Install

```bash
python -m venv .venv
.venv\Scripts\activate
pip install otree
```

### Run locally

```bash
otree devserver
```

Then open the local oTree URL shown in terminal.

Default admin credentials in `settings.py` are currently:

- username: `admin`
- password: `admin`

For production or shared environments, move credentials/secrets to environment variables.

## Notes and Current Limitations

- Matching and social interaction are modeled ex-post; no real-time wait pages/group synchronization are used.
- Topic text and option labels are currently placeholders (`topic A-E`, `Option A/B`); update these before deployment.
- A `guessing_game` is not a separate app in the current implementation; beliefs are collected inside `room_composition`.

## Suggested Next Documentation Additions

- Add a reproducible data dictionary mapping each exported column to app/page/round.
- Document payoff computation rules for both practice and real rounds.
- Add a short deployment section (Heroku/server process, environment variables, admin hardening).