💾 Archived View for r.bdr.sh › gemlog › 1724692197354 › recently-read-collective-problem-solving.gmi captured on 2024-09-29 at 00:02:49. Gemini links have been rewritten to link to archived content

-=-=-=-=-=-=-

Recently Read: Collective Problem-Solving of Groups Across Tasks of Varying Complexity (2020)

Authors: Almaatouq, Abdullah & Yin, Ming & Watts, Duncan

The article @ ResearchGate

For this paper, the authors ran an experiment trying to understand the conditions under which groups perform better than individuals, and what attributes best predict their performance: skill, social perceptiveness, or cognitive style. The problem selected is a constraint satisfaction problem where N students have to be placed in M rooms, following Q constraints. Complexity is adjusted by varying N, M and Q.

Based only on the problem selected, the results here have limited application: For higher numbers of N, M and Q, i'd categorize the problem as what cynefin [1] calls complicated. I believe this will also bias the results towards expertise, and sure enough: the study concludes that expertise is the biggest indicator on performance, in this case skill being measured by participants' ability to solve the same type of problem.

[1] The Cynefin Framework

I was disappointed by the limited definition for cognitive style it focuses on two strategies: whether participants first optimized room utility or constraints. I would have liked to see testing on how participants analyze the problem before attempting to solve it, as I come with a lot of bias on this [2].

[2] This study by Larkin & Reif was a big part of how processes were shaped in one of the first product teams I was part of, and I thought it worked really well. I've carried a lot of those methods with me.

Finally before talking about the results: I don't know enough about "Reading the Mind in the Eyes" as a measure of social perceptiveness, so I can't make any judgement on conclusions in that area.

The findings show that both social perceptiveness and skill have a positive and significant effect, but skill outperforms by an order of magnitude. I believe this is due to the problem type: I'm not sure whether the same would hold for complex or chaotic problems. My prediction is that problems where cause and effect are not clearly known will benefit a lot more from diversity than skill.

Skill diversity has a negative effect on score but not efficiency. I'm not surprised by this one either since it looks at only at how good participants are at this type of problem and not a variety of problem solving skills. For this variable I'd like to see how individual scores are affected in groups: Do individuals with lower skill improve their skills better with groups or individually? Do they do so better with groups with high social perceptiveness? Do they do so better with groups of varied skill levels, or with comparable peers?

Overall, I didn't find the task used particularly interesting, but I think it does reinforce an approach for complicated problems: groups of experts will be able to analyze the problem more effectively, and provide a better response.