How artificial intelligence can minimise uncertainty in group decisions

The method improves group decisions and ranks alternatives by means of an ideal solution.

This article is based on research by Núria Agell

Is there such a thing as a perfect decision? Probably not, especially when it has to be taken in a group setting with many divergent opinions. Findings by Esade Professor Núria Agell take artificial intelligence a step further and provide answers that take us closer to reaching the best solution in group decisions.

Published in the Journal of Information Science, the researchers have developed a mechanism that minimises uncertainty and balances divergence of opinions in the search of the ideal answer. "The method could be very useful in terms of resolving decision-making problems and speeding up group decisions in the coming years," says Agell.

The method could speed up group decisions in the coming years

One of the benefits of this method is that it does not require communication between the experts in the committee, and therefore minimises the risk of having experts who are more persuasive and may lead to final decisions that are not necessarily the best ones. 

"The experts don't need to meet; they just need to introduce their evaluation of alternatives into the system. This way the method ensures that consensus is reached in a more impartial manner."

Capturing uncertainty

The first step in the study was to capture the lack of precision among some of the experts and then aggregate the resulting answers for each of the alternatives presented to solve the problem. 

The second step was to assign linguistic labels for each of the options and measure the distance between each of the alternatives to find an artificially constructed ideal solution. "The system is capable of capturing and measuring uncertain answers by using different degrees of precision and then, based on all the data collected, create the ideal alternative."

3 solutions for real-case scenarios

Grounded in mathematical calculations, the study analysed opinions in three fields in which group decisions are crucial:

  1. Civil engineering projects
  2. Doping control laboratory accreditation
  3. Improving retailing performance

The first real-case scenario was to choose the better of two options for the construction of a subway line in Barcelona. The method analysed all the evaluators' judgments through linguistic labels and identified the first alternative – adopting a tunnel diameter of 12m with the stations included within the tunnel – as the best solution. "This was the solution adopted by experts in the case of Barcelona underground, which highlights the accuracy of our method in this real-life case," says Agell.

Subway tunnel
Photo: Ricardo Gomez Angel/Unsplash

The second case was linked to the accreditation process that laboratories involved in doping controls have to go through and which require a large panel of evaluation experts. "Reaching an agreement in the expert committee takes a long time. Our method shortened this time and proved that the level of quality was the threshold defining the maximum distance value to optimum for a laboratory to be accredited." 

In the third real case, the goal was to detect the features that describe a firm's performance. The case was based on a retail firm in Taiwan and examined information on 44 of the firm's features, provided by 84 expert managers. "The method led to two different rankings of the features that were ratified by an advisory committee and which demonstrated the experiment's accuracy."

The method also revealed the advantages of allowing evaluators' judgments with different levels of precision and not requiring an average of the judgments.

The practical results of these findings are encouraging and could prove useful if applied to real-life managerial decision making. The research group is already working on the next steps to make this happen: "We are developing web-based software to allow group decision-making meetings anytime and anywhere."

Agell says that the group's next research challenge will be to explore machine-learning techniques that will allow the updating of information for decision-making and come up with a system that is capable of adapting to changes in real-world settings.

All written content is licensed under a Creative Commons Attribution 4.0 International license.