Some practical questions for an exponential age
Photo by Andrea Cappiello on Unsplash
Some questions for high school students (and their teachers) about algorithms, which are the underlying logic of machine learning, deep learning, and artificial intelligence in general:
Should an algorithm be used to calculate the risk of recidivism for criminals?
How (if at all) should judges rely on that algorithm to determine how long someone is sent to jail?
How often should that algorithm be "retrained" using updated bodies of data?
What should happen when the algorithm updates itself in ways that increase or decrease the recommended sentencing for people who have already begun to serve their sentence?
Should an algorithm be used to characterize the utility of early parole for prisoners already in the system?
Does the public have a right to know how any such algorithms are built (e.g., transparency about what variables it considers, and what weight it assigns to those variables)?
Who should determine the categories of data and the origins of the data sets to train these algorithms?
These are not idle questions. We already use algorithms to shape sentencing. Cathy O'Neil has reported on this in the must-read Weapons of Math Destruction. And last week, former Facebook engineer / current Harvard JD candidate Ellora Thadaney Israni shared similar concerns in the New York Times.
We teach students about algorithms (primarily through the study of functions) in high school Math. Do those students understand that our criminal justice system relies on them?
In an age of exponential change, things move faster and faster. Paradoxically, this means that some of our thinking needs to slow down: we must devote time to considering how we want to shape our society using our exponential technologies.
Because if we don't, our exponential technologies will do that shaping for us.
***
Do you know someone who would find the Ed:Future blog worthwhile reading? Please let them know that they can subscribe here.