technews-editor@acm.org
Date: Mon, 17 Apr 2017 12:16:09 -0400 (EDT)
Princeton University 13 Apr 2017 via ACM TechNews 17 Apr 2017
Researchers at Princeton University have demonstrated how machines can be
reflections of their creators' biases. They determined common
machine-learning programs, when fed ordinary human language available
online, can obtain cultural prejudices embedded in the patterns of wording.
"We have a situation where these artificial intelligence [AI] systems may be
perpetuating historical patterns of bias that we might find socially
unacceptable and which we might be trying to move away from," warns
Princeton professor Arvind Narayanan. The team experimented with a
machine-learning version of the Implicit Association Test, the GloVe
program, which can represent the co-occurrence statistics of words in a
specific text window. The test replicated the broad substantiations of bias
found in select Implicit Association Test studies over the years that relied
on human subjects. Coders might hope to prevent the perpetuation of
cultural stereotypes via development of explicit, math-based instructions
for machine-learning programs underpinning AI systems.
https://orange.hosting.lsoft.com/trk/click?ref=znwrbbrs9_6-13472x2118efx072995&