Chaos

The Chaos blocker research programme

Last updated: 19 October 2026

The question we are working on is simple to state and hard to answer: how can you tell that work is stuck when nobody says so? This page describes how we are going about it, what we have built so far, and how researchers can take part. The programme is early and we have not yet published results.

How the study is set up

Every signal Chaos uses to suspect a blocker is written down as a testable cue: a kind of wording, a pattern in reporting, a missed date, a stretch of silence. Each cue has a starting weight, and several weak cues combine into one score. When the score is high enough Chaos asks the owner one private question. Their answer is the label.

That gives us a dataset in which every question we asked is stored with the cues that were present at that moment and what the person said in reply. It is the raw material for measuring which cues actually predict a real blocker, and for replacing hand-set weights with measured ones.

What we will and will not do with data

  • Research data is de-identified: no names, phone numbers, message text or per-person records. It carries the cues present, the outcome, and timing.
  • Workers are told in plain language what is stored and who sees it, and can ask to see or delete their data.
  • We do not score individual pace or productivity, and no cue is allowed to become a ranking of people.
  • Hypotheses are written down, with the metric and the minimum sample, before they are tested.

Questions we think are open

  • Hedging and deferral. Do repeated hedges (“I’ll see”, “trying”) predict a blocker better than a single one, and how does that differ across English, Luganda and Swahili?
  • Slipping dates. Estimates are known to run optimistic (the planning fallacy). Does a passed due date with no update separate a blocked item from one that is merely late?
  • Silence. Silence is ambiguous: leave, no signal, or reluctance to share bad news. Which combinations of silence and context justify asking, and which are better left alone?
  • Reluctance to report. The “MUM effect” describes a reluctance to pass on undesirable news. Which wording of a question makes the honest answer easier?
  • Asking well. How many questions a week before a helpful prompt becomes a nag, and does that threshold differ by role?

Background reading

  • Rosen, S. & Tesser, A. (1970). On reluctance to communicate undesirable information: the MUM effect. Sociometry, 33(3).
  • Edmondson, A. (1999). Psychological safety and learning behavior in work teams. Administrative Science Quarterly, 44(2).
  • Morrison, E. W. & Milliken, F. J. (2000). Organizational silence. Academy of Management Review, 25(4).
  • Buehler, R., Griffin, D. & Ross, M. (1994). Exploring the planning fallacy. Journal of Personality and Social Psychology, 67(3).

Take part

We are looking for researchers and students in psychology, organisational behaviour, linguistics and statistics, particularly those who work on East African workplaces and languages. If you would like to propose a hypothesis, review our method, or help with Luganda and Swahili wording, write to chaos@yugiautomate.com.

The measure at the centre of the work is explained on the blocker latency page.