Description
A checklist of eight observable signatures (sudden jump, multimodality, inaccessibility, divergence, hysteresis, anomalous variance, divergence of linear response, critical slowing down) used together to decide whether a behaviour changes through a genuine tipping point rather than a smooth acceleration.
Many behaviour changes look sudden in a time series: a user abandons a product, a habit forms, an attitude flips. The book points out that a fast but continuous acceleration and a true phase transition can produce identical curves. What separates them is whether the in-between states are stable. Gilmore's catastrophe flags are a set of phenomena that a cusp (two stable states with an unstable region between) predicts and a smooth process does not. No single flag is decisive, but combinations are: the book singles out inaccessibility plus hysteresis as convincing. Some flags need only cross-sectional data (multimodality, divergence of the distribution as a splitting variable such as involvement increases), others need repeated measures (sudden jumps, critical slowing down) or experimental control of a variable (hysteresis, inaccessibility). For UX research the flags give a disciplined way to test claims such as 'adoption tips once a team reaches a critical mass' or 'trust collapses suddenly after errors', instead of reading tipping points into any steep curve. The choice of which flags to test depends on what you can measure and manipulate.
Attributes
- Discipline
- Systems & Complexity
- Loss type
- Latent
- Stage
- Discovery, Post-launch
- Data
- Quantitative
- Effort
- Weeks
Steps
- Define the behavioural variable and state the two hypothesised stable modes, the unstable middle and the candidate control variables (one shifting the balance, one splitting the modes, such as involvement).
- Start with cross-sectional data: test the distribution of the behavioural variable for multimodality with a mixture model; if it is unimodal, reconsider the hypothesis.
- Test divergence: check whether the distribution becomes bimodal only at high values of the proposed splitting variable.
- Where you have dense repeated measures, test for sudden jumps with change-point methods and look for rising variance, autocorrelation or slower recovery before transitions.
- Where you can manipulate the normal variable, run an ascending and descending sweep to test hysteresis, and try to hold participants in the middle state to test inaccessibility.
- Tabulate which flags were found and judge the combined evidence; then consider fitting the cusp model directly.
Sources
- van der Maas, H., Complex Systems Research in Psychology (2024)
Also known as
- Gilmore's catastrophe flags
- Phase transition flags
- Cusp flags
Categories
- Strategy & Collaboration
- Systems & Complexity