A starting point for discussion
The idea
The author describes a possible fairness benefit: automated profiling might avoid some bias that people may not notice in their own decisions. But the concern is not profiling itself. It is targeting people arbitrarily rather than on legitimate grounds, meaning relevant reasons for targeting them, such as being genuinely more likely than others to commit the crime in question. The author presents reduced bias as a possibility, not a guarantee.
Place in the reasoning
AI assessment · Connected to reasoning
The possible fairness benefit of automated profiling is explicitly conditional and balanced by the distinction between legitimate and arbitrary targeting; this bears on trade-offs in using algorithmic systems.
Adapting to computer systems or resisting them can both carry risks
It supplies a possible fairness benefit that may be weighed against risks, while warning that reduced bias is not guaranteed.
What this means and why it matters
What this means
Profiling can be unfair when it targets groups arbitrarily, without relevant grounds. The passage says targeting should instead rest on relevant grounds, such as a person being genuinely more likely than others to commit the crime in question. Automated systems might avoid some bias that people may not notice in their own decisions, but they are not guaranteed to do so.
Why it matters
To judge whether a system is fair, it matters both whether it reduces bias and whether it targets people on legitimate grounds, meaning relevant reasons for targeting.
Conditions & assumptions
Source coverage: The full text of the research paper written by the author, read from a copy hosted on another site; not the final version formatted by the publisher. Passage 25 distinguishes arbitrary discriminatory targeting from targeting on legitimate grounds, meaning relevant reasons for targeting, then describes a possible advantage of automated systems. Passage 26 adds that automated systems can also reproduce bias.
Questions this connects to
AI-suggested relationships. Authors have not necessarily responded to one another.