This panel discussion will bring academia, business and government perspectives on the need for an ethical framework that governs core values and principles for the use of Artificial Intelligence (AI) and Big Data in enterprises. As AI becomes more embedded in enterprise decision making, safeguards need to be in place to periodically ensure that these so-called black box algorithms adhere to the core values of an enterprise. Business leaders have the challenge to be trustworthy in the age of artificial intelligence. They need a new ethical framework that redefines how they use AI in their products and services and innovate safely and confidentiality. This framework will address important ethical issues around fairness, privacy, accountability, interpretability, confirmation bias, and transparency. It will help business leaders in striking a balance between protecting their enterprise IP vs. transparency. There are plenty of examples of unfairness perpetrated by unchecked usage of AI in enterprises – both commercial and government, which Cathy O’ Niel terms as Weapons of Math Destruction (Weapons of Math Destruction by Cathy O’ Niel, Broadway Books, ISBN: 978-0553418835) and Virginia Eubanks equates with Automating Inequality (Automating Inequality by Virginia Eubanks, St. Martin’s Press, ISBN: 978-1250074317). Following are the questions that the panel will try to discuss: a. how do we ensure that the AI models use only those proxy variables that accurately and objectively measure the underlying predictor variables; b. How do we ensure that the data the AI models use to test and train themselves are a true representative of the entire population those models are going to affect; c. How do we ensure that the AI models are set to minimize false positives and false negatives in such a way that the innocent and qualified may not be treated adversely; d. How do we ensure that are algorithms are not second-guessing the very same variables which our institutions and organizations are legally barred Balakrishnan et al. CIO Panel on Ethical Framework for Big Data Proceedings of the Fourteenth Midwest Association for Information Systems Conference, Oshkosh, Wisconsin May 21-22, 2019 2 from using explicitly in decision making; e. How do we ensure that we, as owners and renters of these algorithms, are providing transparent and objective feedback to all stakeholders on how these models function. Then we other issues about the deployment of these systems – a. how can we ensure that the data will not be used for any other purpose not intended or stated during data collection; and b. How do we ensure that these AI models are delivered with a “user manual” on how to feed the data, how to flag inaccurate and missing data, how to interpret the outcomes, when to use and when not to use these models. And above all, how do we ensure that we not only have a sound ethical framework but more importantly how to ensure that the framework is religiously followed, and timely reviewed for its relevance and effectiveness. The unchecked deployment of algorithms in consumerization of insights actioned from AI and Big Data is increasing efficiency at one hand but is also magnifying the inaccuracies and unfairness that existed before these systems were designed and implemented. It is important to analyze how these algorithms exaggerate human biases pertaining to motivated reasoning and confirmation biases among others. An ethical framework is needed to check these algorithms since unfairness is something that no business, institution or society can afford to have.
Balakrishnan, Murali; Bansal, Gaurav; Cagigal, David; Mehta, Raman; and Thiel, Todd, "Panel Discussion: CIO Panel on Ethical Framework for AI & Big Data" (2019). MWAIS 2019 Proceedings. 6.