A huge thank-you to my collaborators and advisors, who made this joint work by @csspenn.bsky.social and @mitsloan.bsky.social possible! This work is a huge part of my PhD dissertation, and I'm so excited and proud to share it after 5+ years! 🎓 20/20

TL;DR: Not all tasks are created equal, and now we have a way to measure the difference.

📚 READ OUR PAPER HERE: doi.org/10.1287/mnsc... 19/20

Whether you're a manager or a social scientist, we hope the Task Space can help you move from puzzling local observations ("my team failed at this task!") to understanding how they generalize ("when and why are some teams more likely to fail?"). 18/20

This can be a tool for navigating a range of task-dependent challenges that organizations face: which tasks should be delegated to AI, and which benefit from human teams? Which tasks need specialists, and which can be done just as well by anyone? 17/20

Rather, our finding is a call to *put the task front and center in organizational decisions*. The Task Space is a structured recipe for quantifying how tasks relate to each other. 16/20

This isn't a recommendation to *never* assign open-ended tasks to teams. Here, our metric of success is performance; but organizations care about other outcomes too. For example, working on open-ended problems as a team can be rewarding for learning and building rapport. 15/20

Other creative tasks are more open-ended, like coming up with a story or ad. These are the tasks where groups tended to under-perform individuals. 14/20

These task characteristics cut across traditional categories. Some creative tasks, for example, are well-defined, like generating valid words from a list of letters (it's easy to check your work with a dictionary!). 13/20

For more open-ended tasks, groups were prone to getting stuck in conflict or settling on a sub-optimal solution. 12/20

So what makes a task good for teams? In our data, teams excel on well-defined tasks: those with verifiable answers, divisible subparts, and minimal uncertainty. Here, groups can most effectively divide-and-conquer the work. 11/20

Critically, the Task Space makes it possible to predict when two heads are better than one; our 24D space captures nearly half of the variance in team performance. In other words, if you want to know whether it's worth it to build a team, the task is HALF THE STORY. 10/20

An Illustration of the Task Space. Left: Our labeled set of tasks can be thought of as a 102 tasks × 24 dimensional matrix in which each task is represented as a row vector for which each element (column) is a dimension of the task. Right: Each row vector in this matrix can also be mapped to a point in 24-dimensional space. This representation makes the Task Space easily amenable to linear algebra-based analysis (e.g., finding similarity between vectors, clustering, and sampling).

📍Introducing the Task Space: a map that locates tasks in a 24-dimensional space of theory-backed features. 🗺️ 9/20

What if, instead, we could capture task distance in a manner similar to semantic spaces from AI/ML? Just as we can measure the "distance" between words using word vectors, could we use vectors to represent how similar or different two tasks are? 8/20

But all hope isn't lost. The problem with task categories is that they're too coarse; they don't truly measure the distance between one task and another. 7/20

Frustratingly, this variation doesn't map onto common "task categories." Some creative tasks were among those where groups performed best; others were among the worst. When we used categories to predict group advantage, the models explained a shocking 0% of the variance. 6/20

A figure illustrating Heterogeneity in Group Advantage Across Conditions. Each point represents the observed group advantage at the level of an experimental condition, which is a tuple of task (y-axis) × level of complexity (color/line thickness; low complexity is represented by a thin green line, medium complexity by a mid-thickness purple line, and high complexity by a thick orange line) × group size (point shape; small, three-person groups are represented by circles and large, six-person groups are represented by crosses). Tasks are grouped by the experimental wave in which they appeared (top three facets). Error bars represent the analytical 95% confidence intervals (1.96 * 1 standard error) for group advantage in a given condition. Boxes are centered around the mean and show one standard error.

Across tasks, there's massive variation in team performance. There’s variation in moderation too: for some tasks, groups do *better* as the task gets *more* difficult. For others, the opposite is true, and groups do *worse* when things get harder. 5/20

We ran an experiment with 20 diverse tasks, 2 group sizes (3-person and 6-person teams), 3 levels of difficulty (Low, Medium, and High), and over 1,200 participants. On some tasks, groups did 3x worse than the best-performing individuals. On others, groups did 1.6x better. 4/20

The key lies in something we often overlook: the task at hand. 3/20

Groups are hailed for their "collective intelligence." Companies compete to build teams with the best talent. We design our work around teams: brainstorming sessions, huddle rooms. But what if, in some situations, it doesn't make sense to work in a team at all? 2/20

A screenshot of the paper's title and abstract in the journal Management Science. Title of paper: "The Task Space: An Integrative Framework for Team Research." Abstract: Research on teams spans many contexts, but integrating knowledge from heterogeneous sources is challenging because studies typically examine different tasks that cannot be directly compared. Most investigations involve teams working on just one or a handful of tasks, and researchers lack principled ways to quantify how similar or different these tasks are from one another. We address this challenge by introducing the “Task Space,” a multidimensional space in which tasks—and the distances between them—can be represented formally, and use it to create a “Task Map” of 102 crowd-annotated tasks from the published experimental literature. We then demonstrate the Task Space’s utility by performing an integrative experiment that addresses a fundamental question in team research: when do interacting groups outperform individuals? Our experiment samples 20 diverse tasks from the Task Map at three complexity levels and recruits 1,231 participants to work either individually or in groups of three or six (180 experimental conditions). We find striking heterogeneity in group advantage, with groups performing anywhere from three times worse to 60% better than the best individual working alone, depending on the task context. Critically, the Task Space makes this heterogeneity predictable: it significantly outperforms traditional typologies in predicting group advantage on unseen tasks. Our models also reveal theoretically meaningful interactions between task features; for example, group advantage on creative tasks depends on whether the answers are objectively verifiable. We conclude by arguing that the Task Space enables researchers to integrate findings across different experiments, thereby building cumulative knowledge about team performance.

When is it worth it to hire a team, compared to one competent individual?

📢 NEW PAPER (out this month in Management Science!) by me, @mark.whiting.me, @linneagandhi.bsky.social , @duncanjwatts.bsky.social, and @amaatouq.bsky.social! 🧵1/20

If you're at @aomconnect.bsky.social #AOM2025, come check out our panel on NLP in organizations, featuring @dholtz.bsky.social, Sameer Srivastava, @williambrady.bsky.social, @mikeyeomans.bsky.social & me! We'll explore topics including hiring, cooperation, agency, and moral outrage -- don't miss it!

And if you want the tl;dr: 📖 bsky.app/profile/cssp...

A HUGE thank you to the team behind this project — @bstewart.bsky.social, Dean Knox, Yasemin Savas, Naijia Liu, @jdbk.bsky.social, @adamberinsky.bsky.social & others from a huge inter-university collab between Wharton, Princeton, Harvard, MIT, & WashU. I learned so much from this process!

Our paper is out today in @pnas.org! 🎉 In a large-scale experiment on a YouTube-like platform, we find that giving people politically “slanted” video recs doesn’t shift beliefs or viewing behaviors.

In other words, online filter bubbles may not be as polarizing as we think…

Emily Hu starting off the last session at #SJDM with a bang! Text analysis for JDM

Next up @xehu.bsky.social @sdpbht.bsky.social @mikeyeomans.bsky.social

#managementstudies #orgstudies #mgmtsky

Can computational tools, like LLM’s, turn segmentation algorithms, and NLP, improve theorizing in management? 💻💭Join our symposium on 8/12/24 @ 11:30 AM at the Hyatt Regency Chicago: Burnham Room! Feat. me, Burint Bevis, Mohammed Alsobay, Gus Cooney, Laurie Weingart, and Randall Peterson! #AOM2024

I imagined A Land Before Time when papers roamed, but only some survived 🦕🦖📜

haha I just wrapped my batch of AOM reviews and 100% of them used “extant” in exactly this way 😅