Our modern, digital culture feeds off visual data, captured in photos from various sources. This data, when coupled with powerful technologies like facial recognition, can pose a privacy concern. But, technologies today see beyond our physical appearances. They capture our behaviors, language, reactions, and relationships, leading to some serious implications.

Digital tech can provide insights about our behaviors, preferences, interactions, etc.

Example: Use of algorithms in criminal justice. They calculate a potential offender's likelihood to reoffend (recidivism risk). This is then used to suggest prison sentences.

Irony: Longer sentences have been shown to increase recidivism rates.

Concern: These tech systems have shown bias towards ethnic groups and economic classes, contributing to a cycle of discrimination.

🔮 Predictive Knowledge Limitations

Both human and machine predictions have their limitations.

Questions arise

• How do we know if machine predictions are more reliable than humans?
• What knowledge is necessary to evaluate the validity and neutrality of these algorithms?
• Who decides the assumptions these models should be based on?

💡 Real World Example: Consider the US prison system. People of color often receive higher recidivism risk scores leading to longer sentences. This results in fewer opportunities upon release and higher recidivism risk scores for others in their community. A tragic feedback loop!

🔗 Want to dive deeper? Check out Cathy O’Neil's talk at Google about “Weapons of Math Destruction”.