How expectedparrot/edsl works
EDSL is a computational social-science and market-research toolkit. Researchers define question types, optionally combine them with agent personas, models, and scenario data, then run surveys or experiments to collect structured responses and analyze them as datasets. The graph emphasizes that design-to-results path, including survey logic, interview execution, model inference, and results. It also preserves documented Coop cloud capabilities and scenario data preparation. Edges are limited to README statements or the supplied source excerpts and index references; unsampled subsystems are represented only where documented.
Main entry points: survey.py Survey and logic · interview_task_manager.py Interview tasks · jobs.py Job execution · results.py Structured results
Components 21
Research design
- Question types
edsl/questions/question_base.py - Survey and logic
edsl/surveys/survey.py - Skip and stop rules
edsl/surveys/rules/rule_manager.py - Scenario data
edsl/scenarios/scenario.py - Scenario sources
edsl/scenarios/sources - Agent personas
edsl/agents/agent.py
Run execution
- Interview tasks
edsl/interviews/interview_task_manager.py - Job execution
edsl/jobs/jobs.py - Response cache
edsl/caching/cache.py
Model inference
- Model selection
edsl/language_models/model.py - Inference registry
edsl/inference_services/inference_service_registry.py - Inference services
edsl/inference_services/inference_service_abc.py - Model providers
edsl/inference_services/services
Results and analysis
- Structured results
edsl/results/results.py - Dataset analysis
edsl/dataset/dataset.py
Cloud collaboration
- Human validation
edsl/coop/coop.py - Coop cloud API
edsl/coop/coop.py - Remote response cache
edsl/caching/remote_cache_sync.py
Other components
- Researcher
- Researcher
- LLM provider
Connections 27
- Researcher → Question types: defines
- Researcher → Survey and logic: composes
- Researcher → Agent personas: configures
- Researcher → Model selection: selects
- Researcher → Scenario sources: imports data
- Scenario sources → Scenario data: creates scenarios
- Survey and logic → Question types: contains questions
edsl/surveys/survey.py - Survey and logic → Skip and stop rules: delegates rules
edsl/surveys/survey.py - Survey and logic → Scenario data: parameterized by
edsl/surveys/survey.py - Survey and logic → Agent personas: runs with
edsl/surveys/survey.py - Survey and logic → Model selection: runs with
edsl/surveys/survey.py - Survey and logic → Job execution: creates jobs
edsl/surveys/survey.py - Job execution → Interview tasks: runs interviews
- Interview tasks → Question types: creates tasks
edsl/interviews/interview_task_manager.py - Interview tasks → Skip and stop rules: follows dependencies
edsl/interviews/interview_task_manager.py - Interview tasks → Inference services: requests answers
- Model selection → Inference registry: resolves service
- Inference registry → Inference services: selects service
edsl/inference_services/inference_service_registry.py - Inference services → Model providers: dispatches inference
edsl/inference_services/services/google_service.py - Model providers → LLM provider: sends prompts
- Job execution → Response cache: uses cache
- Job execution → Structured results: produces responses
- Structured results → Dataset analysis: supports analysis
- Structured results → Researcher: presents results
- Researcher → Human validation: validates with people
- Researcher → Coop cloud API: shares research
- Coop cloud API → Remote response cache: supports remote caching