Essai 2 · 18 November 2026
From a focus dot to a warranted claim
A focus table cannot tell us whether researchers studied factors in isolation. Reading the original papers reveals relationships—but their evidential strength varies sharply.
The first essay reconstructed Oraee and colleagues' 62-row table and exposed the gap between counted research focuses and a claim about fragmentation. This attempt enters six of the original studies. It finds explicit cross-factor relationships in singly coded papers and in the blank row, while also finding a printed quantitative inconsistency and a technical relationship that should not be mistaken for evidence about collaboration.
Inspect the passage-to-claim chain
Five categories follow Oraee's framework: Context, Process, Task, Team and Actor. The original focus marks remain untouched. The links below are new, explicitly contestable coding of relationships in particular passages.
Does the least-ready partner determine how the team exchanges information?
In Papadonikolaki and colleagues’ case E, uneven BIM capability accompanied ad hoc exchange. Case B also included a subcontractor without BIM readiness, yet the partnership used sophisticated distributed exchange. That subcontractor bought BIM drafting support. These cases rule out the simple inference that one unready participant fixes the team’s exchange pattern. They give a project team a more precise question to investigate: how does the partnership obtain the missing capability?
The authors used observation, documents and interviews. Recording how a team exchanged information and attributing the exchange pattern to a particular cause make different demands on that evidence. Select row 47 and compare the exchange arrangement with the readiness proposition under the same evidential test.
One relation can defeat a reading; it cannot establish a field
Adamu is marked only Actor in the review's table, yet the paper reports skill-dependent delegation of modelling tasks. That is enough to reject an inference from one focus dot to absence of studied relationships. Ochieng and Price's completely blank row is also not evidence of relational absence. It describes participants' explanations involving interpersonal conduct, trust and joint action, although its transport to BIM networks remains open.
Papadonikolaki and colleagues go further into the actual organisation of BIM work. Their cases link meetings, agreements, experience and information exchange. Their negative observation—that the collaboration patterns did not simply track the number of supply-chain activities—also matters: counting practices is not the same as understanding their configuration.
None of this establishes that the review's broader diagnosis of fragmentation is false. A literature can contain several relationships and still lack a well-specified, tested system connecting all five categories. The defensible correction is to separate focus coverage, relation presence, evidence for a mechanism and transport to the intended setting.
A closer read also weakens a tempting source
The Abuelmaatti and Ahmed survey reports relationships between participant profiles and their ratings of guidance. It therefore cannot simply be described as studying Process in isolation. But its Table 5 buy-in cells, as printed, do not reconcile with the sample.
| Group | Printed cells | Row sum | Group size used elsewhere |
|---|---|---|---|
| Architecture | 8, 19, 27 | 54 | 27 |
| Engineering | 5, 12, 17 | 34 | 17 |
| Construction | 16, 4, 20 | 40 | 20 |
The cells sum to 128, twice the reported 64 respondents. The last cells equal the group totals; they may be misplaced totals, but that is an interpretation, not recovered data. The PDF page was visually checked to distinguish a printed inconsistency from an extraction error. This specific cross-tabulation cannot support a quantitative conclusion as printed. It does not by itself invalidate every other result in the paper. Open the original Table 5.
What counts as a relationship is itself a research decision
Amann and Borrmann offer a useful negative case. Their relationships between geometric parameters are computational structure, not automatically links among collaboration antecedents. Those passages remain excluded. A different passage requires agreement between experts and vendors over exchanged programs. That can reasonably be read as an Actor–Process condition, but it is an untested governance requirement and its mapping into this review's categories is debatable. The permissive switch admits precisely that extra interpretation.
Abrishami's proposed tool-to-capability link is clearer as a relationship, but its survey records expectations. Accepting the relationship code must not promote favourable perceptions into demonstrated capability gains. The interactive claim selector therefore changes the verdict even when the source and its focus marks remain identical.
How the audit was conducted
Adamu and Papadonikolaki were known leads from the preceding source review. Before the detailed reading of the other four for this audit, the contrast rule selected the first two table rows, the first later single-Process technical method, and the only all-blank row: Abrishami, Abuelmaatti, Amann and Ochieng. All six original papers were retrieved; none was substituted after an access failure.
The unit was an explicit relation with named constructs and a pinpoint passage. Co-mention alone did not count. Each coding records the original focus, proposed relation, supporting observation or argument, strongest objection and scope. Directed content analysis supplies this shift from category counts to propositions and relations, and its warning about confirming prior expectations. Hsieh and Shannon, Three Approaches to Qualitative Content Analysis (2005), pp. 1281–1283.
This is one AI analyst's documented source coding with an adversarial rereading, not independent human agreement. The six papers are a purposive challenge sample, not an estimate of prevalence among all 62 papers. Original observations, interviews and statistical data have not been replicated. The arithmetic check and preservation of the original focus marks are independently executable; the meaning of a passage still requires scholarly judgement.
The complete six-source audit · Arithmetic and provenance checks · Unchanged original focus transcription · Oraee et al. (2017), §§2.3, 5–6
Selections are temporary. This reader saves no annotations or altered source codes. Reload restores the opening view.