Are You an Ethics Champion? Five Questions to Ask Before Employing Predictive Analytics in Practice
1. Would it be appropriate to intervene today based on a prediction about the future?
It may not be okay (or legal) to hold people accountable for actions they have not taken, decisions they have not actually made, or rules they have not yet broken. In some cases, even those much more likely than average to have an outcome in relative terms are unlikely to have that outcome in real terms.
2. How would you safeguard against letting structural inequalities that may be hiding in some data systems seep into your practice?
Some data may reflect inequalities in how the data were collected (like greater visibility and reporting of families living in crowded apartment buildings, for example), or in past governmental practices (such as more concentrated policing of some communities compared to others), or in normal human error (such as cognitive biases or poor data collection and entry protocols).
3. What would be the consequences of a predictive analytics tool that leads to over-intervening (or under-intervening)?
Not all governmental interventions are entirely benign. Some individuals or families could be harmed by interventions based on “false positives.” But neither are all interventions unwanted impositions. Proving extra services could be welcome in some cases.
4. Would the use of predictive analytics increase transparency and accountability or diminish it?
Highly technical approaches, “black box” tools, trade secrets, or a lack of data sharing may weaken how government is held accountable to the public.
5. Would using predictive analytics move your mission forward, or would using predictive analytics imply altering your mission?
Technology, tools, and programs should not shape values. Values should shape technology, tools, and programs.