Decision Making in the age of AI: Leveraging AI to improve and ensure Decision Quality.
The SAIL Research Group is focused on how we adapt and leverage the use of AI for Decision Making and to promote collaborative Decision Dialogue in order to make better decisions. The research group is particularly interested in how AI can assist in dealing with the interdisciplinary challenges associated with wicked and strategic problems in society, as well as studying the influence AI will have on the decision sciences overall.
Decisions are the actions we can control to ensure that we realize the value and the benefits we envisage as possible in a situation. To make a good quality decision we first need to frame the problem or opportunity and then consider our alternatives and criteria for making a choice. A key step is to gather meaningful and reliable information that will enable the proper framing and comparative analysis work.
Decision Quality requires involvement of stakeholders to address all relevant issues, uncertainties, and ambiguities, as well as to ensure that we have not excluded potentially important perspectives, and that we are thinking with sound reasoning and logic.
Correct use of data and information is essential to decision making, coupled with clarity on the values and preferences of each stakeholder. At the same time, we need to be aware of biases and potential conflicts. Throughout the decision process, this awareness of biases and preferences of both stakeholders and subject matter experts require that we engage and consider ways to elicit information and ensure that biases and fallacies are recognized and resolved. Examples of bias (and fallacy traps) include but not limited to: anchoring, motivational, confirmation, evidence, saliency, heuristic, crowd wisdom or groupthink, availability, sunk cost and planning fallacy, self-interest, overconfidence and overoptimistic bias, risk or loss averse, halo, historical or “business as usual” lock in traps.
Data and information are essential through the decision making process, from understanding the situation, to framing the problem or opportunity, to the use of models to analyse the alternatives and to inform the decision maker of the insights gained. At each point, we can make use of the opportunities that AI can offer with the ability to provide new insights, avoid human biases in using data, and quickly create insights about a complex situation. Of course, we need to be aware of the potential pitfalls and traps AI may bring and ensure that equal quality is put into how the AI models are used. Biases could be made more prominent with incorrect use of AI, or be better resolved with a thoughtful, guided design of the models and queries used.
The main aim of this research group is to consider how we can leverage AI to help support decision making, how we can carefully craft prompts and use of the tools and techniques available in decision science to make sure we properly use AI to help us discover the problem and potential solutions while allowing for the divergence on the problem at hand and convergence on an optimal decision. Integration of AI into decision making may help avoid frame blindness and assist in reframing the situation based on new information. It may also surface preferences and perceptions from stakeholders we may not have been cognizant of without the AI assist. AI may eventually enable high quality tools for developing alternatives with a sufficient range and divergence before they are used to help analyse and converge on a choice based on the preferences and criteria we have established. The goal is to arrive more quickly and confidently at a shortlist where we can use the insights to discuss trade offs and reach an optimal decision. AI should add considerably to the decision quality of the process by a set of tools and processes to help check the reasoning, logic and transparency throughout the process before we select a preferred option and commit resources to it.
The long history of decision sciences has been a struggle with the cognitive biases of humans, gaps caused by frame blindness, lack of information from all stakeholders, and the time required to gather and make sense of conflicting information. Research into the integration of AI with decision sciences will aim to leverage the potential speed, breadth of data considered, and the synthesis capabilities of AI to address these limitations and move to a new, higher level of decision quality. In conclusion, this research aims to explore what guidelines and AI models we should leverage in order to debias and deal with the problems of framing and analysis of decisions and ultimately what ways AI can improve Decision Making in wicked, multi-stakeholder situations with interdisciplinary dimensions.