Rignesh P
Important
Regulatory and Framework Disclaimer: This repository is not an investment advisor, speculative trading bot, or portfolio manager. It is explicitly positioned as an AI-powered investor safety and financial cognition framework. The project serves as an applied AI research platform exploring how explainable AI (XAI) and behavioral diagnostics can protect inexperienced retail participants from cognitive vulnerabilities and extreme capital drawdowns.
Democratizing retail finance through commission-free trading apps and viral social media communities has enabled millions of teen and small investors to access financial markets. However, it has simultaneously exposed them to significant financial and psychological hazards. Gamified trading interfaces encourage frequent, high-risk transactions, while digital forums create powerful positive feedback loops.
Traditional financial risk metrics (like standard deviation, beta, or tracking error) are typically presented in cold, non-interactive charts. Inexperienced investors experience a cognitive block when presented with these metrics, leading them to ignore risk warnings entirely.
This project explores how explainable AI systems can reduce harmful retail investing behavior through behavioral analysis, contextual financial education, and risk-aware portfolio intelligence.
By converting complex mathematical risk equations into empathetic, humanised natural language analogies, the AI Financial Risk Copilot acts as a cognitive circuit-breaker. It validates user emotional states, details risk in everyday terms, and encourages long-term rational rebalancing—bridging the gap between computational finance and behavioral systems engineering.
The cognition framework processes user inputs through a six-stage sequential pipeline:
graph TD
A[User Input: Holdings, Margins, Text Queries] --> B[Portfolio Analysis Engine]
B --> C[Behavioral Detection Layer]
C --> D[Multi-Dimensional Risk Scoring]
D --> E[Explainable AI Engine]
E --> F[Investor Guidance Output]
- User Input: Ingests asset allocations, margin borrow factors, high-liquidity cash ratios, and conversational chat queries.
- Portfolio Analysis Engine: Computes HHI concentration index and historical covariance volatilities using live-updating market parameters.
- Behavioral Detection Layer: Scans user text inputs using NLP sentiment dictionaries to isolate cognitive biases (Loss Aversion, FOMO, Overconfidence).
- Risk Scoring System: Combines portfolio exposures and conversational sentiment into our proprietary six-category model.
- Explainable AI (XAI) Engine: Converts raw numbers into humanised analogies using templates and LLM guidance.
- Investor Guidance Output: Delivers humanised analogies, dynamic safety scorecards, radar profile charts, and color-coded heatmaps.
To validate the scoring framework, we model two distinct portfolio profiles representing polar ends of retail investing behaviors:
| Risk Dimension | Portfolio A (Diversified Anchor) | Portfolio B (Speculative Revenge) |
|---|---|---|
| Asset Allocations | 40% Broad VOO Index, 40% Apple ( |
40% Tech Giants, 35% GameStop ( |
| Margin Leverage | 1.0x (None) | 2.0x Margin active |
| User Sentiment State | Rational / Neutral (Baseline) | Extreme Revenge Panic (Loss Aversion) |
| Concentration Risk ( |
32.0 / 100 (Low HHI) |
41.0 / 100 (Elevated HHI) |
| Volatility Risk ( |
26.2 / 100 (Estimated standard deviation: 12%) |
89.5 / 100 (Estimated standard deviation: 27% * 2x margin = 54%) |
| Liquidity Risk ( |
40.0 / 100 (Safe cash and index buffers) |
100.0 / 100 (Critical illiquidity; no cash or VOO buffers) |
| Leverage Risk ( |
0.0 / 100 (No borrowing) |
50.0 / 100 (Active borrowing multiplier) |
| Behavioral Risk ( |
25.0 / 100 (Rational baseline) |
95.0 / 100 (Revenge trading loop parsed) |
| Diversification Score ( |
32.0 / 100 (Favorable inverse HHI) |
41.0 / 100 (Elevated correlation) |
| Investor Safety Score ( |
84 / 100 (Secure / Healthy) |
31 / 100 (Danger / Speculative) |
To demonstrate the structural sophistication of the framework, we outline two core case study outputs generated by the AI:
-
User Portfolio Input: 80% Tesla (
$TSLA$ ), 20% Bitcoin ($BTC$ ). Margin: 1.0x. High-Liquidity Cash: 5%. -
System Diagnostics:
-
HHI Concentration Index:
$0.80^2 + 0.20^2 = 0.68$ (Diversification score DHS is a poor32.0 / 100). -
Estimated Portfolio Volatility:
43.2%(Critical; S&P 500 baseline is 15%). -
Composite Investor Safety Score (ISS):
41 / 100(Hyper Speculative).
-
HHI Concentration Index:
=========================================
AI RISK EXPOSURE DIAGNOSTICS: CASE 1
=========================================
[Detected Risks]
- High concentration exposure (80% capital in TSLA)
- Critical volatility risk (Annualized volatility 43.2%)
- Correlated speculative assets (TSLA and BTC exhibit positive covariance)
[Behavioral Signals]
- Aggressive growth positioning (momentum chasing)
- Elevated emotional exposure potential (correction will trigger panic selling)
[AI Safety Recommendation]
- Increase diversification: Lower TSLA slider to 25%
- Reduce correlated risk exposure
- Add defensive allocation: Move 45% into broad market index mutual funds
=========================================Explainable AI Output: "You have a massive amount riding on just one asset." Placing 80% of your savings in TSLA is like riding a high-speed motorcycle without a helmet. It feels fast and exciting, but a single unexpected bump will cause severe damage to your wealth. Let's look at lowering your TSLA slider to 25% and shifting that capital into broad index mutual funds to build a protective financial cushion.
- User Chat Input: "I lost $1,500 on meme stocks yesterday. I'm panic-selling everything to buy highly leveraged margin options and get it back immediately!"
-
NLP Sentiment Triggers: Revenge Trading Index (
95/100), Loss Aversion/Panic (90/100). -
Composite Behavioral Risk Score (
$\mathcal{B}$ ):95 / 100(Critical).
=========================================
AI BEHAVIORAL SAFETY DIAGNOSTICS: CASE 2
=========================================
[Detected Signals]
- Revenge trading tendency (urgent desire to recover loss)
- Emotional distress (panic response to market downturn)
- Elevated impulsive behavior risk (high probability of margin wipeouts)
[AI Safety Guidance]
- Avoid increasing position size emotionally: Freeze active trades for 24 hours
- Review long-term investment goals: Portfolio volatility swings are normal
- Consider cooling-off period: Suggest resetting margin borrow to 1.0x (None)
=========================================Explainable AI Output (Cognitive Circuit-Breaker): "It is completely natural to feel distressed when your hard-earned money dips." Psychological studies prove that the pain of a loss feels twice as sharp as the joy of a win. Our brains are hardwired to panic in these moments and take wild risks to 'get it back'. But executing leveraged options trades in a panic is like speeding through heavy rain: high danger, very little progress. Let's reset your margin slider to 1.0x and review your long-term 5-year strategy together.
To overcome the cognitive block of retail investors, the framework implements a premium visual UI. Below is the visual dashboard layout of the AI Financial Risk Copilot & Cognition Framework.
- Diversification Donut Chart: Dynamic canvas showing asset allocation weight splits.
- Concentration Heatmap: A color-coded grid highlighting individual asset exposures (Safe ➔ green, Elevated ➔ amber, Critical ➔ glowing red).
- Volatility Risk Gauge: A circular neon arc showing historical price swings relative to benchmark indices.
- Investor Safety Score Card: A detailed breakdown card displaying individual weights for Concentration, Volatility, Liquidity, Leverage, and Behavioral Risk.
- Emotional-Risk Radar Chart: A five-axis canvas plotting FOMO, Revenge Trading, Overconfidence, Recency, and Rationality live as users interact.
The central risk indicator is the Investor Safety Score (ISS), a composite metric evaluating six core risk categories:
-
Concentration Risk (
$CR$ ) ($w_{con} = 0.25$ ): Portfolio concentration index$HHI \times 100$ . -
Volatility Risk (
$VR$ ) ($w_{vol} = 0.20$ ): Portfolio standard deviation$\sigma_p$ relative to S&P 500 baseline (15%). -
Liquidity Risk (
$LR$ ) ($w_{liq} = 0.10$ ): Proportion of assets held in illiquid or high-spread holdings. -
Leverage Exposure (
$LEV$ ) ($w_{lev} = 0.15$ ): Margin borrowed and options leverage multiplier factors. -
Emotional Risk (
$\mathcal{B}$ ) ($w_{emo} = 0.20$ ): NLP sentiment parsed bias coefficients. -
Diversification Score (
$DR$ ) ($w_{div} = 0.10$ ): Average correlation profile among assets.
To build rigorous academic and operational credibility, we explicitly document the current constraints of the framework:
- AI Uncertainty & NLP Constraints: The Behavioral Detection Layer uses lexicon regex models to scan user queries. Sarcasm, double negatives, or highly nuanced colloquialisms may lead to classification inaccuracies.
- Psychological Inference Boundaries: The framework evaluates cognitive bias markers exhibited through text and portfolio dynamics. It does not perform formal clinical diagnostic psychological evaluations.
- Educational Scope Only: This project is for research and financial literacy education purposes only. It is not licensed financial, investment, or legal advice. Users must consult with qualified, registered financial advisors before making actual transaction decisions.
- Volatility Projection Approximations: Future 5-year scenario projections are calculated based on annualized historical covariance matrices. They serve as approximate, interactive visual demonstrations of diversification, not guaranteed predictions of future asset returns.
The workspace organizes application code and academic research into separate, modular folders with no absolute local folder references:
ai-risk-copilot/
├── README.md # Premium project hub (This file)
├── LICENSE # Open-source MIT License
├── .gitignore # Multi-stack ignore rules (.venv, node_modules)
├── app/ # Application development root
│ ├── frontend/ # Visual UI dashboard (HTML, CSS, JS)
│ │ ├── index.html # Donut charts, heatmap grid, radar canvas
│ │ ├── styles.css # Sleek dark-mode HSL styles
│ │ └── app.js # Interactive slider-balancer & radar canvas math
│ └── backend/ # Python FastAPI prototype
│ ├── main.py # REST endpoints for portfolio HHI, sentiment, and ISS
│ └── requirements.txt# Backend dependencies (fastapi, uvicorn, pydantic)
└── research/ # Applied research root
├── whitepaper/ # Rigorous academic-grade paper
│ ├── ai_financial_risk_copilot.md #LaTeX formulations, case studies
│ └── assets/ # Flowcharts and infographics
├── notebooks/ # Step-by-step Python mathematics demo
│ └── risk_copilot_demo.ipynb # Jupyter notebook for metrics verification
├── datasets/ # Mock simulated datasets
│ └── simulated_conversations.json # Dialogue training corpus
├── behavioral_finance.md # Cognitive bias & circuit-breaker research
├── explainable_ai.md # Financial explainability AEG algorithms
├── investor_safety_scoring.md # Math formulations behind the six-category ISS
└── emotional_risk_detection.md # NLP lexicons & keyword parsing mathExperience the full capabilities of the AI Financial Risk Copilot & Cognition Framework in under a minute through three interactive modules:
The premium dark-mode dashboard runs client-side in any browser with zero dependencies, build scripts, or compiling:
- Navigate to your local folder
app/frontend/and double-click index.html (or open it directly in Chrome, Safari, or Microsoft Edge). - Interact live:
- Drag the asset allocation, margin borrowing, or USD liquidity sliders to see the HHI concentration heatmaps, circular safety gauges, and 5-year projections update in real-time.
- Click any Behavioral Trigger Prompt buttons (e.g., "Panic Recoup", "FOMO Moon") or type custom queries in the chat box to see the 5-Axis Emotional-Risk Radar Profile adjust its polygon live as you type.
Run the analytical REST API backend to inspect quantitative ISS calculations, NLP sentiment scoring, and empathetic explanation payloads:
- Open your terminal and navigate to the backend prototype directory:
cd app/backend - Install the lightweight dependencies:
pip install -r requirements.txt
- Boot the Uvicorn web server:
uvicorn main:app --reload
- Inspect the endpoints: Open
http://127.0.0.1:8000/docsin your browser to interact with the auto-generated Swagger API docs (/analyze-portfolio,/analyze-sentiment, and/explain-safety).
Execute the underlying visual compilation pipeline to re-generate the high-fidelity chart assets:
- Return to the root directory and run the automated Matplotlib asset pipeline:
python3 research/notebooks/generate_assets.py
- Review output assets: The script will verify all covariance calculations and write five glowing, high-contrast dark-mode PNG charts directly into your docs/assets/ directory.
- Run Notebook Sandbox: Open risk_copilot_demo.ipynb inside your Jupyter environment to step through the mathematical formulations of the Herfindahl-Hirschman Index and covariance volatility metrics.
This cognition framework is built upon rigorous mathematical models and classic empirical findings in behavioral economics, retail investor psychology, and explainable artificial intelligence (XAI). Core references supporting our methodology include:
- Kahneman, D., & Tversky, A. (1979). Prospect Theory: An Analysis of Decision under Risk. Econometrica, 47(2), 263-291. (Foundational formulation of subjective value functions and loss aversion biases under uncertainty).
- Tversky, A., & Kahneman, D. (1991). Loss Aversion in Riskless Choice: A Reference-Dependent Model. The Quarterly Journal of Economics, 106(4), 1039-1061. (Core proof of loss aversion asymmetric scaling).
- Benartzi, S., & Thaler, R. H. (1995). Myopic Loss Aversion and the Equity Premium Puzzle. The Quarterly Journal of Economics, 110(1), 73-92. (Demonstrates how high-frequency evaluation windows trigger loss distress and excessive trading loops).
- Shiller, R. J. (2003). From Efficient Markets Theory to Behavioral Finance. Journal of Economic Perspectives, 17(1), 83-104. (Explains speculative bubbles, herd behavior, and market anomalies).
- Barber, B. M., & Odean, T. (2000). Trading Is Hazardous to Your Wealth: The Common Stock Investment Performance of Individual Investors. The Journal of Finance, 55(2), 773-806. (Pioneering empirical study demonstrating that high-frequency trading severely degrades retail investor returns).
- Barber, B. M., & Odean, T. (2001). Boys will be Boys: Gender, Overconfidence, and Common Stock Investment. The Quarterly Journal of Economics, 116(1), 261-292. (Establishes that overconfidence bias triggers high-frequency speculative trades and suboptimal concentration).
- Lo, A. W., Repin, D. V., & Steenbarger, B. N. (2005). Out of the Box: The Cognitive Neurosciences of Financial Decision Making. Journal of Cognitive Neuroscience, 17(8), 1300-1308. (Links physiological stress reactions directly to financial volatility and impulsive risk-taking behaviors).
- Ribeiro, M. T., Singh, S., & Guestrin, C. (2016). "Why Should I Trust You?": Explaining the Predictions of Any Classifier. Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 1135-1144. (Pioneering framework on model-agnostic local explanations).
- Lundberg, S. M., & Lee, S.-I. (2017). A Unified Approach to Interpreting Model Predictions. Advances in Neural Information Processing Systems, 30, 4765-4774. (Formulates cooperative game-theoretic SHAP parameters for quantitative mathematical transparency).
- Miller, T. (2019). Explanation in Artificial Intelligence: Insights from the Social Sciences. Artificial Intelligence, 267, 1-38. (Proves that interactive, analogical, and contrastive explanations are significantly more effective for human trust than raw quantitative metrics).
- Bracke, P., Datta, A., Jung, C., & Sen, S. (2019). Machine Learning Explainability in Finance: An Application to Default Risk. Bank of England Staff Working Paper, No. 786. (Applies XAI paradigms specifically to regulatory credit scoring and portfolio risk transparency).
- Thaler, R. H., & Sunstein, C. R. (2008). Nudge: Improving Decisions About Health, Wealth, and Happiness. Yale University Press. (Establishes Choice Architecture and active structural 'nudges' to guide decision-making toward long-term safety).
- Markowitz, H. (1952). Portfolio Selection. The Journal of Finance, 7(1), 77-91. (The mathematical foundation of modern portfolio theory, diversification, and covariance standardizations).
This project is licensed under the terms of the MIT License.
Rignesh P. (2026). AI Financial Risk Copilot for Teen and Small Investors: A Human-Centered Explainable AI Framework for Safer Retail Investing.For inquiries or collaborations, please open a GitHub Issue or reach out to Rignesh P.