Bias in Elections

Upcoming Policy Event

AAAS Forum on Science & Technology Policy
Washington, D.C., June 21 – 22, 2018.
From AAAS: “The annual AAAS Forum on Science and Technology Policy is the conference for people interested in public policy issues facing the science, engineering, and higher education communities. Since 1976, it has been the place where insiders go to learn what is happening and what is likely to happen in the coming year on the federal budget and the growing number of policy issues that affect researchers and their institutions.”

Follow-up on  the April 1 Policy Post: Experiments on FaceBook Data

 US organizations and individuals influence voters through posts in social media and analysis (and misanalysis) of publicly-available data. Experimentation has been reported on the use of FaceBook data to show techniques that can be used to change elections (Nature, volume 489, pages 295–298 (13 September 2012)). Particularly, the authors looked at data during the 2010 US Congressional elections and showed how to affect voting. They report “results from a randomized controlled trial of political mobilization messages delivered to 61 million Facebook users during the 2010 US congressional elections. The results show that the messages directly influenced political self-expression, information seeking and real-world voting behaviour of millions of people. Furthermore, the messages not only influenced the users who received them but also the users’ friends, and friends of friends.”

For more information and analysis, see Zoe Corbyn’s article “Facebook experiment boosts US voter turnout.”

FaceBook, Google, and Bias

Current events involving FaceBook and the use of data they collect and analyze relate to issues addressed by SIGAI and USACM working groups on algorithmic accountability, transparency, and bias. The players in this area of ethics and policy include those who are unaware of the issues and ones who intentionally use methods and systems with bias to achieve organizational goals. The issues around use of customer data in ways that are not transparent, or difficult to discover, not only have negative impacts on individuals and society, but they also are difficult to address because they are integral to business models upon which companies are based.

A Forbes recent article “Google’s DeepMind Has An Idea For Stopping Biased AI” discusses research that addresses AI systems that spread prejudices that humans have about race and gender – the issue that when artificial intelligence is trained with biased data,  biased decisions may be made. An example cited in the article include facial recognition systems shown to have difficulty properly recognizing black women.

Machine-learning software is rapidly becoming widely accessible to developers across the world, many of whom are not aware of the dangers of using data contain biases.  The Forbes piece discusses an article “Path-Specific Counterfactual Fairness,” by DeepMind researchers Silvia Chiappa and Thomas Gillam. Counterfactual fairness refers to methods of decision-making for machines and ways that fairness might automatically be determined. DeepMind has a new division, DeepMind Ethics & Society that addresses this and other issues on the ethical and social impacts of AI technology.

The Forbes article quotes Kriti Sharma, a consultant in artificial intelligence with Sage, the British enterprise software company as follows: “Understanding the risk of bias in AI is not a problem that that technologists can solve in a vacuum. We need collaboration between experts in anthropology, law, policy makers, business leaders to address the questions emerging technology will continue to ask of us. It is exciting to see increased academic research activity in AI fairness and accountability over the last 18 months, but in truth we aren’t seeing enough business leaders, companies applying AI, those who will eventually make AI mainstream in every aspect of our lives, take the same level of responsibility to create unbiased AI.”