| 🎯 Goal | Examine the ethical challenges of automated decisions made with prescriptive analytics |
| 🧠 Approach | Frame Linear Integer Programming (LIP) risks and propose concrete safeguards |
| 📊 Focus | Over-optimisation & data privacy in optimisation models, via a worked case study |
| 📈 Delivery | Presentation deck + Excel LIP examples |
Prescriptive analytics doesn't just describe or predict — it decides. Linear Integer Programming (LIP), solved efficiently with the Branch & Bound algorithm, automates choices under integer constraints across retail, manufacturing, and healthcare.
The tension: A model tuned purely for efficiency or profit can quietly harm the people it schedules, prices, or serves.
This project asks how to keep powerful optimisation fair, private, and accountable — so automation builds trust rather than liability.
| Element | Detail |
|---|---|
| ⚙️ Technique | Linear Integer Programming (LIP), solved via Branch & Bound |
| 🏬 Domains | Retail, manufacturing, healthcare decision automation |
| Over-optimisation (profit over wellbeing); misuse of sensitive data | |
| 🧪 Worked example | Retail staff-scheduling scenario with Excel LIP calculations |
RESPONSIBLE OPTIMISATION WORKFLOW
──────────────────────────────────────────────
1. Model the decision as an LIP problem
2. Identify where pure optimisation causes harm
├─ Over-optimisation → staff wellbeing
└─ Data misuse → privacy breach
3. Inject ethical safeguards:
• Welfare constraints (fairness)
• Anonymisation + consent (privacy)
• Audit trails + monitoring (accountability)
• Multi-objective formulation (balance)
4. Re-solve → balanced, compliant decision
5. Weigh trade-off: added compute & compliance cost
──────────────────────────────────────────────
Four ethical principles, a challenge-to-safeguard map, and the retail-scheduling case study. The case narrative is illustrative of the scenario presented (representative — no exact metrics claimed).
- ⚙️ Over-optimisation is the central risk — models chasing efficiency or profit can neglect employee wellbeing.
- 🔒 Data privacy is easily compromised — sensitive employee and customer data may be used without proper safeguards.
- 🟢 Welfare constraints restore fairness — encoding wellbeing directly into the model prevents exploitative outcomes.
- 🔵 Anonymisation + consent protect privacy — without blocking the optimisation itself.
- 🟣 Multi-objective optimisation balances efficiency and responsibility, rather than optimising profit alone.
- 💡 Ethics is a business advantage — safeguards reduce legal liability, build trust, and support long-term sustainability, at the cost of added computational and compliance overhead.
| Stakeholder | Benefit of ethical safeguards |
|---|---|
| 👷 Employees | Protected from exploitative, efficiency-only scheduling |
| 🤝 Customers | Data handled with consent and anonymisation |
| 🏢 Business | Lower legal liability, stronger brand & stakeholder trust |
| 📉 Trade-off | Added constraints raise computation and compliance cost |
| Category | Tools |
|---|---|
| Optimisation | Linear Integer Programming (LIP), Branch & Bound |
| Analysis | Microsoft Excel (LIP examples & calculations) |
| Concepts | Multi-objective optimisation, data anonymisation, welfare constraints |
| Deliverable | Presentation deck + supporting spreadsheet |
Ethical Considerations in Prescriptive Analytics/
├── 📁 assets/
│ ├── 🎨 banner.svg # Repository banner
│ └── 📊 dashboard.svg # Prescriptive analytics dashboard
├── 📁 data/
│ └── 📊 Integer Linear Programming.xlsx # ILP examples & calculations
├── 📁 docs/
│ └── 📄 Final Presentation.pptx # Challenges, safeguards & case study
└── 📝 README.md # Project overview
Heta Chavda — Data Analytics | Optimisation | Data Ethics
Group presentation with Krishna Patel, Pavan Naik & Jayram Karan
⭐ Found this useful? Give it a star!