Hierarchical Task Analysis vs. Cognitive Work Analysis

Summary

Comparative methodology paper from human-factors / ergonomics literature. Compares the two dominant task-analysis frameworks — HTA (Hierarchical Task Analysis; Annett et al. 1971, Stanton 2006) and CWA (Cognitive Work Analysis; Vicente 1999, Rasmussen 1986) — across theoretical underpinning, methodology, and contribution to system design. Empirical case study: a military helicopter Mission Planning System (MPS) software tool, analysed with both approaches.

Relevant for the BPM wiki primarily as a reference source for observation-based elicitation protocols — HTA's data-collection mix (observation + interviews + walkthroughs + document review) and its iterative observe → draft → refine-with-SME cycle are directly portable to process-model discovery.

HTA — how it works

Theoretical core: decompose an overall goal into sub-goals, operations, and plans. An operator (human or machine) is "required to do X" to achieve a super-ordinate goal. Roots in scientific management (Taylor, Gilbreths) via Miller, Galanter & Pribram (1960) Plans and the Structure of Behaviour. HTA developed in the 1960s specifically to describe cognitive aspects of task attainment (previously task methods focused only on physical/observable behaviour).

Procedure (Stanton 2006, Fig 1):

State the overall goal.
State subordinate operations.
State the plan (sequence/conditions for operations).
Check adequacy of redescription → revise if not OK.
Consider first/next suboperation → recurse if further redescription required, else terminate.
Repeat until all operations terminated.

Data collection techniques — not a single method:

Observation (shadowing operators)
Questionnaires / structured interviews with SMEs
Walkthroughs (talk-aloud while doing the task)
User trials
Documentation review

Iteration: "high degree of reiteration, often with appropriate SMEs, before the analysis is completed to a satisfactory level." Single-pass is never adequate.

Granularity is load-bearing: in the paper's case study, HTA went to button-press level (e.g. "click on intervisibility drop-down menu, click right mouse button to activate mapping menu, click mouse on map to drop waypoint"). For error prediction (SHERPA) this granularity is appropriate; for BPM discovery it is far too fine and would obscure the end-to-end flow.

Task-decomposition categories (Kirwan & Ainsworth 1992)

For each low-level task step, HTA captures:

Category	What to note
Initiating event/cue	What triggers this step
Information requirements	What the operator needs to know
Interface features (controls / displays)	Where the information is and how it is accessed
Actions required	Physical/cognitive actions
Decisions required	Choices the operator must make
Complexity / difficulty	Is it routine, error-prone, cognitively heavy
Outputs	Result of the step
Feedback	How the operator knows it worked

This template is directly usable as an observation field-note structure for BPM discovery — per observed activity, capture these 8 fields.

HTA vs CWA — when to use which

	HTA	CWA
Nature	Normative (how it does/should unfold)	Formative (how it could unfold)
Granularity	Fine (button-press)	Coarse (functional)
Strength	Design modification, error prediction, training design	First-of-a-kind design specification, coping with unanticipated events
Phase	Later (evaluate/refine existing system)	Early (idea forming / concept)
Output	Goal-subgoal hierarchy + plans	Abstraction hierarchy + decision ladder + contextual activity template

Paper's verdict: they are complementary. Apply CWA early (formative, specification) and HTA later (normative, refinement). Both need multiple SME iterations.

Utility for BPM discovery (our angle)

Observation-protocol template → the 8-field decomposition table above.
Iteration discipline → observe → draft hierarchy → validate with SME → revise, analogous to Dumas's Interview → Modeling → Validation cycle (sources/2018-dumas-fundamentals-of-bpm-ch5-discovery §5.2.2).
Granularity warning → HTA's button-press case study illustrates exactly what Dumas §6.1 warns against ("put document on fax machine"). Keep BPM discovery at activity or subprocess level.
Multi-method default → no serious task analysis uses observation alone; observation + interview + document-review is the standard combo — aligns with Dumas §5.2.4's "mix methods" recommendation.

Connections

concepts/hierarchical-task-analysis — concept page.
methods/process-discovery-methods — observation method.
sources/2018-dumas-fundamentals-of-bpm-ch5-discovery — Dumas §5.2.1 observation treatment (much thinner than HTA).
syntheses/qualitative-discovery-method-selection-matrix — observation candidate in the matrix.
syntheses/interview-structuring-for-process-models — iteration cycle parallels.
References not ingested: Annett et al. 1971 (HTA origin); Kirwan & Ainsworth 1992 A Guide to Task Analysis; Embrey 1986 (SHERPA); Vicente 1999 Cognitive Work Analysis.