Battling algorithmic bias

system that is using such algorithmic assessments. Algorithms also are used to serve up job listings or credit offers that can be viewed as inadvertently biased, as they sometimes utilize end-user characteristics like household income and postal codes that can be proxies for race, given the correlation between ethnicity, household income, and geographic settling patterns. The New York Times in July 2015 highlighted several instances of al-gorithmic unfairness, or outright discrimination. It cited research conducted by Carnegie Mellon University in 2015 that found Google's ad-serving system showed an ad for high-paying jobs to men much more often than it did for women. Similarly , a study conducted at the University of Washington in 2015 found that despite women holding 27% of CEO posts in the U.S., a search for " CEO " using Google's Image Search tool returned results of which just 11% depicted women. A 2012 Harvard University study published in the Journal of Social Issues indicated advertisements for services that allow searching for people's arrest records were more likely to come up when searches were conducted on traditionally African -American names. For their part, programmers seem to recognize the need to address these issues of unfairness, particularly with respect to algorithms that have the potential to adversely impact protected groups, such as those in specific ethic groups, religious minorities, and others that might be subject to inadvertent or deliberate discrimination. " Machine learning engineers care deeply about measuring accuracy of their models, " explains Moritz Hardt, a senior research scientist at Google. " What they additionally need to do is to measure accuracy within different subgroups. Wildly differing performance across different groups of the population can indicate a problem. In the context of fairness, it can actually help to make models more com-HAVE become an integral part of everyday life. Algorithms are able to process a far greater range of inputs and variables to make decisions, and can do so with speed and reliability that far exceed human capabilities. From the ads we are served, to the products we are offered, and to the results we are presented with after searching on-line, algorithms, rather than humans sitting behind the scenes, are making these decisions. However, because algorithms simply present the results of calculations defined by humans using data that may be provided by humans, machines, or a combination of the two (at some point during the process), they …