Ethical considerations in decision-making

Demystifying Decision Science

Akshay Swaminathan

PD Soros Fellow at Stanford University School of Medicine

The responsibility of Data Science

Data Science empowers decisions across industries
Ethical use is essential for trust and fairness

Case Study: Optum

Optum's algorithm unintentionally favored healthier patients
- Designed to predict care needs, it prioritized those with higher healthcare spending
Bias emerged from flawed data assumptions
- Using spending as a proxy for need led to racial disparities, disadvantaging those with less access to care

Fairness

Privacy and security

Transparency and reproducibility

Bias in machine learning models

bias data.png

Bias in = Bias out

Machine learning models reflect the biases and gaps in the data they are trained on

Fairness can be measured

Evaluate model performance across different groups to assess fairness

Protecting data

Decision scientists and organizations must ensure secure and responsible data handling

Involves:

Secure infrastructure with updated protections
Restricted access: only minimum data needed per role
Consent-based data collection for personal data

Why it matters:

Breaches reduce trust
Can lead to lawsuits
Damage reputation and profitability

Transparency

Ensures clarity and accountability

Report data sources, cleaning steps, analysis methods, collected metrics, and uncertainty levels

Reproducibility

Reproducibility builds trust and credibility

Others should be able to repeat the analysis with the same data and methods and get the same results

Ethics at the heart of decision science

Ethical considerations are integral to responsible decision science

Ethics is not optional - it's essential
Prioritize fairness, justice, and benefit
Apply ethical principles across the data lifecycle

Let's practice!

Demystifying Decision Science

Preparing Video For Download...