The impact of individual expertise and public information on group decision-making.

Abstract

Due to growing automatization, and interconnectivity of decision-makers worldwide, global problems with high complexity have to be dealt with by institutions, which not only struggle with financial shortages in doing so, but which also have to cope with changing mindsets, change management, and a lack of experts. Artificial intelligence, and even quantum computers enable new ways of solving problems by supporting experts, but all models are still limited in finding problems.

Human decision-makers seem to be gifted with an invaluable skill: being able to overcome routine, and finding hidden information or in other words: finding problems. This might be linked to latest insights from neuroscience, which show that the human brain manifests the uncertainty problem, i.e. that deception potential is immanent. Even large amounts of data will fail to predict human behavior due to their valance weighting bias. Models running on “Big Data” then interpret deviations from past behavioral patterns as being “irrational” for good reasons. Humans, naturally, can always shift their “macroscopic”, known routine, by indistinguishable inner “microscopic” structural shifts. While this might be commonly known as a “gut-feeling” or “intuition”, this thesis shows by randomized online experiments that performance in non-routine problem-solving is not randomly distributed, but is most likely linked to the decision-maker’s learning environment, working memory capacity and intrinsic motives, such as motivation to invest in reflection time. Curiosity was not confirmed as an intrinsic motivator, but cannot be excluded, as it was only measured by self-reporting questionnaires. Overcoming a known routine or engagement in non-routine problem solving was also measured to be activated by an individual only, when feedback gave good reason to believe that a change of strategy was necessary, i.e. when some decision-maker’s strategy performance was actually influenced by microstructural changes. Experimental results have shown that with growing complexity, performance in overcoming routine pays off. Even when decision-makers invest in thinking time, they solve complex problems with hidden information in less time, and in less work steps. Not every participant has shown high performance in overcoming routine, and depending on country origin, non- routine decision-making performers varied between 5 % and 20 % of all participants. This led to the conclusion that finding experts from a large database using an efficient assessment online tool such as “Curiosity IO” might help solving mentioned challenges of large institutions with low cost, as finding non-routine problem solving performers, using the experiment “Flag Run”, usually lasts only 15 minutes, and can be easily executed by any web browser.

In a second randomized online experiment, participants were not only observed in isolation, but in groups of three. The well-researched cognitive puzzle game “Tower of Hanoi” was used for this purpose. Tower of Hanoi performance predicts executive function, planning performance and performance in overcoming myopic decisions. A three-player version was designed, in which every group participant always had a fair and shared influence over the group’s decision outcome, while not always having a decisive impact on the group’s outcome – in addition, the order of chosen action by each group member mattered. Therefore, even if the entire ruleset was known to an agent, it would still be unable to control the outcome, as intergroup communication between all agents was not possible.

This “algorithm” tried to simulate reality as described above: complex problems of global magnitude make “knowledge” about all direct and indirect influencers, humans and environmental agents alike, a practical impossibility, and therefore, intergroup communication between all decisive agents is restricted, if not impossible, in such real-world problems.

Starting with a “stable”, well-defined problem, being the single-player version of “Tower of Hanoi”, followed by small macroscopic structural shifts in between, the experiment shifts towards a “meta-stable”, ill-defined problem, being the three-player version, also followed by a small macroscopic structural shift in between, which leads to these final ill-defined stages being best described as “instable”. Small macrostructure shifts during stable conditions significantly influenced performance, and was best handled by high levels of expertise. While group performance was not measured, the experiment’s focus was laid upon deviation from routine strategies, i.e. how far participants would deviate from their known routine. A routine, which was mentally built up during the stable, well-defined problem conditions. This routine, when being faced by public information or seemingly chaotic feedback, and under different public information conditions, was then analyzed for behavioral changes. Thus, the focus was laid upon routine deviation in order to simulate “change”.

The thesis’ experimental results are summarized in the following: Public information about environmental change did not influence change in behavior, when feedback gave no reason to believe that this change had an impact on strategy performance during metastable conditions. When change actually had impact on strategy performance, expertise was a good predictor for individuals to change their strategy. During all ill-defined stages, high levels of expertise was followed by less seemingly random choice of strategy, less volatile behavior, and higher chances of an individual to adapt to changes in well-defined circumstances. Throughout all stages, individual expertise was a stronger predictor than public information regarding deviation from routine strategy, i.e. change of behavior. Individual expertise was also a stronger predictor than truthful and deception-free public information, regarding adaption to more effective strategies. In addition, results suggested that potentially discouraging public information, which communicated “lack of control”, led to more seemingly random behavior or loss in motivation to engage the problem with smart heuristics. In addition, expertise in the well-defined domain also predicted behavior in ill-defined problem-solving stages.

Experimental results have shown that expertise was a stronger predictor than public information regarding change in behavior and strategy adaption. The author comes to the conclusion that expertise could be the key factor for global and interconnected problems. Identifying non-routine problem- solving experts by efficient online assessments could lead to less volatile system performance, from which all decision-makers could potentially profit.