The Decision & Risk Analysis track features talks describing effective ways to aid those who must make complex decisions. In particular, the talks reference systematic, quantitative and interactive approaches to address choices, considering the likelihood and impact of unexpected, often adverse consequences. Learn more about confirmed presentations in this track below.
Prevent Disasters: How to Identify and Manage Risks by Learning from Near-miss Events
Although organizations may extract valuable lessons from visible failures, they too often neglect near-miss events—those that occur before a catastrophe—for the early learning opportunities these events can provide. Near-misses are situations where a failure could have occurred except for the intervention of good fortune and are often harbingers of a future failure. Prior research has demonstrated a natural propensity for individuals and organizations to ignore these warning signals because they perceive the near-misses as successes. This presentation will describe more than a decade’s worth of research into identifying near-misses and mitigating risky responses in applications from space missions, natural hazard events, homeland security, commercial aviation, coal mining, and supply chain management.
Cooperative Decision Making with Artificially Intelligent Agents and Humans
The area of human AI teaming often focuses on collaborative operations with distinct task responsibilities for each agent. However, in practice, fully autonomous agents may not be desirable for decision making. Often semi-autonomous systems are required to have a ‘human in the loop’ assisting. A super-agent is a human and AI team who are responsible for the same repeated decision. The goal is to take advantage of the strengths of each individual agent. Using concepts from decision and risk analysis, super-agent thresholds can be established which allow for effective collaboration.
Effectively Managing Cybersecurity Risk: The Rigorous, Revolutionary FAIR Framework
Cybercrimes profoundly affect their victims’ lives and the victimized organizations’ bottom lines. Unfortunately, the shiny tools and “best practice” checklists that comprise cyber risk management are more akin to medieval bloodletting — pernicious practices based on bad reasoning — than to sound analysis and critical thinking. After detailing the deficiencies of current practices, this talk presents Factor Analysis of Information Risk (FAIR), the field’s fastest-growing, most successful cybersecurity risk model — one tested in practice, grounded in sound theory, established as an international standard, and adopted by many of the Fortune 1000. Illustrated with use-cases and examples, you will learn to spot unreliable risk measurements and to help your company understand, measure, and manage cyber risk.
Beyond Bigness: Problems and Solutions for Large Portfolios in Large Organizations
Are bigger portfolios better at diversifying risk?
Unfortunately, size alone does not necessarily create diversification. In fact, it creates risk that is specific to large portfolios. Bigness is driven by success which creates established organizational structure, incentives and portfolio management systems that often direct investment into projects that sustain the existing structure and underestimate risk. Additionally, in many industries, success is driven by a small number of “blockbuster” projects. Over investing in similar projects, underestimating risk, and chasing hot opportunities are systematic problems that can make a large portfolio into a big, risky mess.
Fortunately, we can mitigate these problems and regain the benefits of diversification by acknowledging these challenges and addressing them head-on: by splitting the portfolio appropriately, by developing a robust enterprise-level picture of risk, by aligning incentives with best practices, and by designating a skeptic who is charged with challenging biases and unfounded assumptions.
Decision Analytic Principles for Data Science Leaders
Today, the most valuable companies are those that successfully monetize data. But as the global store of data grows dramatically, companies struggle to hire people with sufficient data science skills and experience. So it is critical that their existing data science teams excel at delivering value.
Because predictive models derive their value from the decisions they inform, data scientists should apply decision analytic principles (e.g., the importance of framing, taking a value focus, pursuing the elements of a high-quality decision) as naturally as they apply data science principles. This talk will demonstrate decision analytic principles on examples of decisions that data science teams encounter, especially those made by the customers of their models.