Beruflich Dokumente
Kultur Dokumente
16 November 2018
Artificial Intelligence
Voigt, Rob, et al. “Language from Police Body Camera Footage Shows Racial Disparities in Officer
www.pnas.org/content/114/25/6521.short.
AI as a Behavior Analyst
It seems like every other day there is news of an unjust case of racial disparity, many a times it
involves law enforcement. Body cams have shown the public the careless slaughter of African Americans
by law enforcement however there are many cases of racial disparity that aren’t accounted for. Who is
keeping track of all the verbal interactions that occur between officers and African American people? It
has been stated that officers treat African Americans with consistent disrespect and researchers at
Stanford University created a computational model that collects evidence to support that assertion. The
computational model creates a “respect index” of sorts by picking up different linguistic features that have
predetermined scores to them. Their results showed that there was indeed a difference in the way officers
addressed white people as compared to African Americans. These results can be used to reconfigure
One thing I found fascinating was the construct behind the model and how there was a huge
importance placed on gathering accurate evaluations of abstract concepts through surveying various
participants and in trying to create technology that thinks like a human, this survey revealed how similar
people are with respect to intuition. The survey asked participants to rate how polite or respectful some
phrases said my officers are and Cronbach’s alpha evidenced there was moderate to high reliability and
consistency among the responses or “data points”. And this brings up another point: math can be used to
validate or invalidate nearly anything and I never realized how far its applicability reached. I did some
external research of the statistical reasoning behind this model and though I haven’t learned how to do
that type of math in school, I still gained clarity about how statistical models are used as a base for
algorithms.
Technology was created to make life simpler and act as a tool for investigation and solving
problems. This scenario represents that perfectly. Social inequity is a plague that has rooted itself deeper
and deeper into America and this innovative approach to evaluating racial disparity in officer interactions
acts as a useful supplement to protests and contentious debates over social reform. There is now hard
proof thanks to this computational model that even accounts for contextual factors such as the severity of
the infraction or the outcome of the stop. And what is even more amazing is the outlook of this model.
Soon it could extend to being able to analyze intonation and “emotional prosody” as the article states it.
This would significantly decrease the likelihood for misinterpreting data on part of the computational
model, especially if paired with video capabilities to factor in facial expressions and body movement. An
interesting point that the article capitalizes on is using the data to evaluate the underlying causes,
surrounding environments, and derive a plan of action for overall amelioration. The data serves a higher