Six research tracks. All hypotheses. None proven.

• R01 Conceptual

  Agent Behavioral Patterns

  Can operational behaviors of AI agents be measured consistently across tasks?

• R02 Planned

  Multi-Agent Coordination

  How do communication patterns, delegation and supervision affect collective outcomes?

• R03 Planned

  Hybrid Workforce Dynamics

  How does collaboration between humans and AI agents affect delivery time, quality and coordination overhead?

• R04 Planned

  Project Outcome Prediction

  Can historical execution data improve predictions of project success, duration and cost?

• R05 Conceptual

  Team Composition Optimization

  Can simulation identify team configurations that perform better under specific constraints?

• R06 Conceptual

  Human–AI Interaction

  How do trust calibration, approval practices, reporting preferences and autonomy levels affect operational results?

Scientific method

1. 1 · Observe

   Collect work events with consent, from declared windows, with their provenance.

2. 2 · Model

   Turn observations into explicit features and simple baseline models first.

3. 3 · Simulate

   Explore configurations with documented assumptions.

4. 4 · Predict

   Register predictions before outcomes are known.

5. 5 · Evaluate

   Score calibration and error; publish what failed as well as what worked.

Four kinds of data, never confused

• Demonstration data

  Illustrates an interface. Means nothing.

• Simulated experimental data

  Output of a stated model under stated assumptions.

• Real collected observations

  Measured, with source and window. None published yet.

• Validated research results

  Survived evaluation against outcomes. None yet.

Experiment Registry

Experiments will be registered here, with hypothesis and analysis plan, before data collection begins.