Consider the following questions about each of the given scenarios:
A company is overwhelmed by the number of applicants that it is receiving for the engineering roles that it has posted.
In an effort to make it easier to hire candidates, a team of data scientists has built a machine learning model that will determine whether or not a candidate should be hired based upon their resume.
Some personally identifiable information, such as the name of the candidate were not considered as a feature of the model.
Judges make decisions about the sentences that criminals receive. One of the factors that judges consider during sentencing is the likelihood of the person to re-offend (recidivism).
Courtrooms have adopted tools designed to eliminate bias in sentencing through the use of artificial intelligence. The history of the criminal can be input into the model. It will then output the likelihood of the person to re-offend.
Demographic information about the criminal is not included in the model.
Child welfare workers are asked to make thousands of decisions related to their work in any given year. This is an overwhelming number of decisions to make - often with imperfect information.
A county’s Department of Health and Human Services has built an algorithm that can aid decision making for its child welfare workers. Each case is given a score that indicates how risky it is based upon the likelihood of the child to be removed from the home within 2 years.
Incidents of potential neglect are reported to the county’s child protection hotline. The reports go through a screening process where the algorithm calculates the child’s potential risk and assigns a score. Child welfare workers then use their discretion to decide whether to investigate.