Forecasting Emergent Phenomena with Human Computer Collaboration
Abstract
Our ultimate goal is to utilize data crowd-sourced from experts towards gaining a quantitative picture of the network of global risks, which in turn will yield some actionable insights. The network by definition has risks of varying complexity, which arguably makes the risk mitigation process more involved for some risks than for others. In such a scenario, our quantification of the relative impacts of different risks could provide an invaluable guide to any cost-benefit analysis involved in the design of policies or strategies aimed at global risk minimization. The next step given the insights provided by our model will be for domain experts to provide tailor-made recommendations for the pertinent risks, such that the likelihood of systemic failures is strongly curbed. The effects of such recommendations can also be tested using our model. Accordingly, the final contribution of the research is to provide policy and organization and team design guidance. Policy guidance would include strategic initiatives that ultimately lead to the control of the network of threats and an early warning system for reliably curbing the disruptive effects of emergent phenomena. For example, we can imagine that strategic guidance on the formation of teams and the interaction between decision making teams and crowds would require new military designs for the selection and training of teams. Early indicators suggest that crowds composed of high diversity, cognitive flexibility and some training in induction can perform best. Because humans tend to be poor at estimating interdependent risks, one area of greater inductive reasoning targeted for skill enhancement might be in the area of identifying the backbone of the network of threats from which cascades are likely to originate. Further, for contributions to organization and team design, we aim to create a methodology, incorporating what we have learned about the best human and machine partnerships that can be generalized to other threat networks on increasingly fine grained levels of specificity such that the model can go from global, to regional, to national levels of analysis. Lastly, we aim to leave a legacy of machine generated bench-marking to calibrate hits rate with the aspiration of continual improvements in prediction and analytics. The technical goals include extending and generalized the existing ARP model to include more than binary state of the modeled risks and to include additional probability distributions of risk beyond the current Poisson distribution. In particular, negative bimodal distribution for rare events will be evaluated and also irregularly cyclic distribution will be defined for cyclic natural disasters. At the same time, we will develop the research designs and tests particularly those designs involving human subject data collection. This sub-goal will involve developing survey questions used to solicit collective intelligence signals about future risks from crowds, and the design of a pilot study and the corresponding methodology in terms of item design, survey design, and appropriate sampling. We will also establish the fundamental properties of the model with focus on rate of convergence of recovery of parameters in constant and individually varying risk parameters. Moreover, we will assess the necessary length of historical data for certain level of confidence for the recovered parameter values. We will complete a pilot study by administering surveys to crowds of executives from his schoolÕs executive education program over two semesters. This population will contain high fraction of high-level executives from all types of firms, both domestically and internationally, including some CEOs of notable companies, as well as physicians, attorneys, and entrepreneurs.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jan 07, 2019
- Source ID
- W911NF1610524
Entities
People
- Boleslaw Szymanski
Organizations
- Army Contracting Command
- Rensselaer Polytechnic Institute
- United States Army