Week 13: Algorithmic Justice

1. Introduction to Algorithmic Justice

• Exploring algorithmic and data justice in relation to the justice system.

• Connection to prior discussions about digital surveillance and the fairness of data use.

• Key Themes:

  • Fair distribution of benefits and burdens from data collection.

  • Importance of algorithms in making justice-related decisions (e.g., sentencing, parole).

2. Algorithms and Decision Making in the Justice System

2.1 Use in Justice

• Algorithms are increasingly used for:

  • Sentencing, probation, parole decisions.

  • Policing via predictive analytics to estimate crime occurrences.

• Discussion on proprietary software in U.S. justice to inform decision-making.

2.2 Human vs. Algorithm Predictions

• Study (2020) comparing accuracy of humans vs. algorithms in predicting reoffending:

  • Initial prediction (5 data points): Humans (64%) vs. Computers (65%).

  • Expanded data set (15 data points):

    • Computers achieved an accuracy of 89%.

    • Humans reached 83% accuracy with training but remained less accurate.

• Raises the question of replacing human judgment with algorithms in the justice system.

3. The Concept of Justice

3.1 Types of Justice

• Broadly classified into three types:

  • Retributive Justice: Punishment based on past actions.

  • Procedural Justice: Fair processes in decision-making.

  • Distributive Justice: Fair allocation of resources and burdens.

• John Rawls’ Justice as Fairness: Influential model in liberal democracies.

3.2 Retributive Justice

• Focused on giving punishment as deserved.

• Emphasizes a backward-looking perspective—punishment for past actions.

• Unique in being a public act, conducted by the state rather than personal vengeance.

3.3 Procedural Justice

• Importance of fair decision-making processes.

• Critical elements:

  • Quality and competence of decision-making.

  • Inclusion of perspectives affected by decisions.

3.4 Distributive Justice

• Fair distribution of benefits and burdens within society.

• Split into two principles:

  • Principle of Merit: Rewards based on what individuals deserve.

  • Principle of Need: Basic rights and protections afforded to all.

3.5 Rawls’ Theory of Justice

• Hypothetical social contract model:

  • Veil of Ignorance: Designing a society without knowledge of personal circumstances to ensure fairness.

  • Maximin Principle: Maximize benefits for the least advantaged in society.

4. Data Justice

4.1 Definition and Importance

• Data justice concerns the interactions between data collection and power dynamics.

• Scrutinizes fairness in the distribution of data burdens and benefits:

  • Who controls data?

  • How are data-related benefits and burdens allocated?

4.2 Examples of Data Justice Injustice

• Digital Divide: Societal inequalities in access to information and communication technologies (ICTs).

  • Digital Inclusion Index findings.

• Targeted Advertisements: Questioning data ownership and commodification of personal data.

5. Algorithms in Criminal Justice

5.1 Understanding Algorithms

• Definition: A set of instructions that converts inputs to outputs.

• Role of algorithms in efficiently processing large datasets and making predictions.

5.2 Type of Algorithms Used

• Algorithms can categorize individuals as likely to offend or reoffend.

• Predictive Policing Algorithms:

  • Examples include spatial-temporal algorithms and individual risk assessments.

• Algorithms used primarily in the U.S., with proprietary software like Compass.

5.3 Strengths and Limitations

• Human oversight cannot be entirely removed; initial biases affect outcomes.

• Potential critique of algorithmic decisions undermining retributive justice.

• Procedural fairness compromised due to lack of transparency in algorithm decision-making (black box problem).

6. Concerns about Algorithmic Justice

6.1 Critique of Distributive Justice

• Algorithms can reinforce existing biases and inequality:

  • Algorithms reflect human-produced data, often marred by systemic bias.

• Example: Bias in the Compass algorithm outputting unequal risk assessments based on race.

6.2 Procedural Concerns

• Lack of explainability may undermine individuals' understanding of adverse decisions.

• Requirement for transparency and understanding of algorithmic processes to maintain procedural justice.

6.3 Conclusion

• Algorithmic justice prompts critical reflection on whether technology can fairly replace human judgment in legal contexts.

• Balance needed in automating justice while ensuring ethical implications are addressed.