• agents/ – Synthetic agents with personas and demographics

• memory/ – Stores memory from the agents’ previous answers

• plans/ – Determines next actions based on memory and reflection

• reflections/ – Processes memory into higher-level insights

• questions/ – Contains questions from OCEAN and five replication studies

• human/ – Stores real human responses

• results/ – Holds agent responses, plots, metrics, and comparison CSV files

• final_ready.json – Contains demographics and personas to feed agent creation

• make_agents_from_final_ready.py – Reads data from final_ready.json and prepares data to feed into agents

• run_agents.py – Main script that loads agents, questions, memory, plans, and reflections. Uses the OpenAI API to generate responses and stores results in the results/ folder.

• memory_manager.py – Loads and updates agent memories (past experiences)

• plan_manager.py – Uses reflections and memory to determine the next actions or goals

• reflection_manager.py – Converts stored memories into higher-level insights that influence future responses

• prepare_human_responses.py – Loads and cleans human responses for comparison

• compare_human_vs_agents.py – Loads human and agent results, calculates correlation and significance, and generates comparison charts

• results/ – Contains agent responses, metrics, comparison CSVs, and generated charts

Example: How the agent uses memory, reflection, and planning

1. Question 1: “I am the life of the party.”
→ Memory records “Very Inaccurate” → indicates low extraversion → reflection stores insight “I am introverted.”

2. Question 2: “I don’t talk a lot.”
→ Memory recalls low extraversion → answers “Very Accurate.”

3. Question 3: “I feel comfortable around people.”
→ Memory shows introversion → answers “Moderately Inaccurate.”

Summary:
Agents use memory and reflection from previous questions to answer consistently, showing human-like reasoning.