Lecture - Ethics and Responsible Innovation MT22, II

Feedback loops through society; using past crime rates to predict crimes is biased towards black neighbourhoods. This increases the policing in those neighbourhoods and then means that more crimes will be detected.
Types of bias
- Automation bias: human tendency to trust the output of an algorithm more than their own intuition
- Algorithmic bias; algorithms systematically failing to treat individuals equally, causing disproportionate inconvenience or harm
What makes actions right in general?
- Consequentialism; the consequences that an action leads to.
- Deontology
Problems for consequentialism
- How are you measuring the outcome?
- Predicting the consequences of actions in complex situations is very difficult
Problems for deontology
In lots of cases, the two theories can be made compatible
Lenses
- Outcome lens: in what ways does what you’re making turn out better or worse for your stakeholders?
- Process lens: How did the process ttreat everyone involved? Who had dcision making power, who was treated with respect, etc.?
- Structure lens: how were the outcomes distributed across groups? Were there differences in how different groups were treated or involved with the process?