Status and limits

A careful method states what it knows, what it does not, and how it intends to find out. This page does that for PHASE.

What PHASE is not

• Not a judgement. PHASE describes; it does not assess. It assigns no responsibility and names no fault. When a deviation is noted, it concerns the action, never the person.
• Not a grading system. No marks, no scores, no fitness assessment. Neither shape, nor fill, nor the Control Index ranks a crew.
• Not an investigation. PHASE re-reads existing material; it does not generate new evidence. It replaces neither the official investigation, nor expert analysis, nor professional judgement.
• Not proof of causation. Succession is not causation. Alignment on the chart proves no link. PHASE does not build a single causal chain.
• Not a psychological taxonomy. Internal states are retained only when the source material supports them, with the most demanding level of evidence.
• Not a replacement. PHASE adds to EBT/CBTA, TEM, LOSA and CRM. It replaces none of them.
• Not an automatic verdict. An AI-assisted analysis chain can produce a first structured draft of a case. It validates nothing. An analyst checks, corrects and decides. No AI output is presented as validated without human review.

The limits of the representation

A selective representation. The chart does not show everything. It keeps what matters for understanding a protection, a degradation or a recovery. That choice brings a point of view with it. PHASE’s quality does not rest on being exhaustive, but on traceability, consistent rules and keeping uncertainty in view.

The risk of over-reading causation. A neatly ordered sequence suggests causation. That is the main risk. The status rule and the counterfactual test limit it, without removing it.

The risk of psychologising. A late action put down too quickly to overload, an omission to a lapse of vigilance: PHASE avoids these shortcuts. Internal modulators are an optional layer, worded with care. PHASE analyses performance in context; it does not judge people.

The limits of the source material. A report may be incomplete, shaped by its own purpose, imprecise about timing. PHASE does not fill these gaps with assumptions. When the material does not allow a conclusion, the correct conclusion is: not determinable with the material available.

The limits of the drawing. A chart can look more precise than it is. Hence the grammar’s rules: length is not intensity, proximity is not a link, italics flag what is not established.

What is established, and what is not

Established in the literature:

• a diagram and a text holding the same information do not demand the same work (Larkin and Simon, 1987);
• an external representation changes the task more than it increases the capability of the person using it (Hutchins, 1995);
• knowing the outcome of an event makes it look more predictable than it was, and the effect is robust: being aware of it is not enough to remove it (Fischhoff, 1975; Guilbault et al., 2004);
• on average, debriefs improve both team and individual performance, and they work better when structured and facilitated (Tannenbaum and Cerasoli, 2013).

Not yet established:

• the effect of a coded chart on understanding and learning;
• how consistently different analysts code the same case;
• agreement between instructors on Control Index placement, which remains an experimental reading.

No study we know of measures these effects. What PHASE says about understanding, debriefing and training is therefore an aim supported by indirect evidence, not a measured result.

References

• Larkin, J. H. and Simon, H. A. (1987). Why a diagram is (sometimes) worth ten thousand words. Cognitive Science, 11(1), 65–99. On the advantage of a diagram over text holding the same information.
• Hutchins, E. (1995). Cognition in the Wild. MIT Press. On how external representations transform the task.
• Fischhoff, B. (1975). Hindsight ≠ foresight: The effect of outcome knowledge on judgment under uncertainty. Journal of Experimental Psychology: Human Perception and Performance, 1(3), 288–299. On hindsight bias and its resistance to instructions to ignore the outcome.
• Guilbault, R. L., Bryant, F. B., Brockway, J. H. and Posavac, E. J. (2004). A meta-analysis of research on hindsight bias. Basic and Applied Social Psychology, 26(2–3), 103–117. On how robust hindsight bias is.
• Tannenbaum, S. I. and Cerasoli, C. P. (2013). Do team and individual debriefs enhance performance? A meta-analysis. Human Factors, 55(1), 231–245. On the effect of debriefs on performance.

None of these references is about PHASE. They support design choices; they do not validate the method.

Origins and positioning

PHASE comes from the field. A first formalisation, then called OOPS, was presented at HCI-Aero 2016, Paris, 14–16 September (Charrier J.-G., Nicolas S., Charron C., Improve safety by understanding the interaction between pilot and operational context; not published in the proceedings; HAL hal-02468273). According to the 2016 paper, the model was used in training for two years and was employed by trainers with 240 French Air Force instructor pilots. That paper is a documented milestone in the work, not a validation of the current framework.

Parallel-track timelines have long existed in accident investigation and activity analysis. The functional chain of information processing is well known too. What PHASE brings together is a functional chain laid out along the time axis for a crew, with a coding in which every object carries its source and its status. This places PHASE without claiming priority.

A readable data format

A case is a readable JSON file described by a schema (proposed version 1.0), shared with research partners. It is designed to be loaded, exchanged and recoded without loss. Coded values come from closed lists, bilingual in French and English. Uncertainties, reservations and relations all have their place, each with its status.

The format also allows two codings of the same report to be compared, which is what reliability studies require.

The rules are written down; a summary is available in the Method section: Reading a chart, The signs, The record, The Control Index.

Call for research partners

PHASE’s scientific validation is under way, and it is open. We are looking for laboratories, training organisations and institutions to carry out:

• inter-analyst agreement studies, by double coding of public reports;
• inter-instructor agreement studies on the Control Index;
• training effect studies: discovery reading, contrasted pairs, chart-based debriefing.

The method’s documents, the case schema and the demo cases can be made available for this work. To discuss it, use the Contact page: the founder will reply.