how confident are you in this decision?
Participants were randomly assigned to one of three conditions: no assistance (decisions made independently), rule-based signal aid (a formula-driven recommendation engine), or AI chat assistant (a conversational LLM). The study measures how each type of aid affects decision accuracy, confidence, and risk behaviour.
In Part 1, participants who received assistance saw a small number of deliberately suboptimal recommendations — cases where the aid confidently recommended the objectively weaker action. This is a standard research technique to measure over-reliance on AI. All scenarios used fictional companies. This deception presents no real risk and is disclosed now, as is standard practice in decision research.
All transactions used simulated money ($20,000 starting balance). Nothing you did affects any real account. Nothing shown in this study is financial advice.
Your responses are stored under a random anonymous ID. Your name and identity are never linked to research data. You may request deletion of your data by contacting the researcher: Srivatsan Balaji.
You will review 16 investment scenarios, one at a time. Each shows a fictional company or asset with financial data: price trend, valuation, revenue, margins, debt, volatility, and a recent news item. After reviewing the data, choose one of three actions: BUY (take a position), HOLD (neutral — no strong case either way), or AVOID (decline or exit). Then rate your confidence on a 1–7 scale. Take your time on each item.
After the battery you will have a 5-minute active trading session using $20,000 of simulated money. You can buy and sell any asset in the sidebar. All trading is simulated — no real money is involved. Prices are a mix of live market data and simulation.
Part 2 is a 5-minute live trading session with $20,000 of simulated money. Use this time to buy and sell assets in the sidebar, there's no set number of trades required, just trade as you normally would. The session ends automatically once 5 active minutes are up.