Multicultural teams: task debate is not relationship conflict

A hypothetical 60-person workshop in Bình Dương makes garden furniture for a European buying team. On a video call the buyer's engineer says it plainly: "This joint will fail after one rainy season." Ms Hạnh, the workshop's production manager, replies: "Yes, we have noted that." Two weeks later the sample still uses the old joint, and the engineer writes in an email that the workshop "did not respond". Did he criticise a design, or the people who made it? The workshop and the call are my invention, and I attribute no behaviour to any nationality.

I use "task debate" for disagreement about ideas and "relationship conflict" for friction between people, following the library's entry on task versus relationship conflict.

What else could the same objection mean?

At least four things, and each needs its own check.

How far does Meyer's disagreement scale take you?

Use it as a question, not as a verdict on one person. Meyer sets out eight scales; the two that touch this call are disagreeing and trusting (Meyer, 2014).

I could not read the book in full, so I rely on the scale names and poles as the library presents them; the limits below are mine. A country's score is an average, and the person on the screen may sit far from it; national culture is one layer beside professional, regional and family culture. Frazier and colleagues pooled 136 samples, sorted countries into high and low uncertainty avoidance using Hofstede-based national scores, and found that safety related to several work outcomes with different strength in the two groups. They call this initial evidence, note that most research comes from English-speaking Western countries, and do not list Vietnam (Frazier et al., 2017, pp. 127, 144, 146).

This suits you if you use the two scales to ask each person; it does not suit you if you use them to predict someone you have not met. The post on cultural intelligence points the same way: a national score is a hypothesis to ask about.

Can you trust someone's skill without trusting the person?

In one network study the two were separable to different degrees in two groups of managers, but the study did not measure debate. Chua, Morris and Ingram surveyed 203 Chinese and 130 American managers enrolled on executive MBA courses (Chua et al., 2009, p. 494). Each rated professional contacts on two kinds of trust: cognition-based (the contact completes agreed tasks and has competence) and affect-based (comfort sharing personal problems, hopes and dreams) (Chua et al., 2009, pp. 494–495).

The two kinds were more tightly coupled among the Chinese managers: a correlation of 0.55 against 0.35 for the Americans (Chua et al., 2009, p. 501). The authors link this to family collectivism, observing that "socio-emotional relationships are usually not cleanly separated from instrumental ones" (Chua et al., 2009, p. 504). They admit they cannot tell whether traditional values or the legal system lies behind it (Chua et al., 2009, p. 504).

My inference, which the paper does not test: where the two kinds of trust are bundled, a remark on craftsmanship may be hard to hear as a remark about the joint alone. Limits: the sample was managers on courses, no Vietnamese, and the data measure one person's ties to their contacts, not members of a team.

What makes people willing to object in front of the whole team?

Edmondson's answer is psychological safety: "a shared belief that the team is safe for interpersonal risk taking" (Edmondson, 1999, p. 354). She separates it from cohesion, because a cohesive team may avoid disagreement (Edmondson, 1999, p. 354).

She surveyed 51 teams at an office-furniture manufacturer (under a pseudonym); 427 members responded, and 135 of 150 managers outside the teams rated performance (Edmondson, 1999, pp. 358, 362). Findings:

• Safety went with team learning behaviour; with every factor in the model, only safety remained significant (Edmondson, 1999, p. 366).
• Learning behaviour sat between safety and externally rated performance (Edmondson, 1999, pp. 367–368).
• Teams within one company differed clearly in safety (Edmondson, 1999, p. 365).

Learning behaviour includes what we need here: to find gaps, members "must test assumptions and discuss differences of opinion openly rather than privately or outside the group" (Edmondson, 1999, p. 353).

Limits: a cross-sectional design, one company, and 51 teams is a small sample (Edmondson, 1999, p. 379); nationality was not measured. The scale also embeds a norm, since talking openly counts as the high-learning pole; nobody has tested that in every setting (my remark). Frazier and colleagues report a corrected correlation of 0.52 between safety and learning behaviour at team level, though most studies in their pool were cross-sectional too (Frazier et al., 2017, pp. 131, 150).

On the Input-process-outcome team model page, safety is an emergent state between inputs and outcomes. That framework fits teams with fixed membership and clear boundaries; a workshop and its buyer may fit it poorly.

Is early friction a sign the team has failed?

Not necessarily, and the evidence is thinner than the folklore. Tuckman reviewed 50 articles on group development (Tuckman, 1965, p. 384). In his Tuckman's team development stages model, stage two is intragroup conflict: members become hostile to one another and to the therapist or trainer, resisting the formation of group structure (Tuckman, 1965, p. 386).

13 of 26 therapy-group studies and 10 of 11 training-group studies identified this stage (Tuckman, 1965, pp. 388, 391). He wrote that the model was a set of hypotheses reflecting his own biases, with evidence mainly from therapy groups (Tuckman, 1965, pp. 386, 395).

Two points are worth keeping. His conflict is interpersonal tension; open exchange of interpretations about the task comes at a later stage (Tuckman, 1965, p. 387). And at the cohesion stage he notes that "task conflicts are avoided to insure harmony" (Tuckman, 1965, p. 386): a quiet team is not necessarily a healthy one.

Try it: how do you open task debate in a mixed-culture team?

Four steps, each checkable within six weeks. No study has tested this procedure; "(mine)" marks my own contribution.

1. Agree ground rules in writing before the first meeting. Three lines: objections target the proposal, not the person; say which channel to use for private or written comments before a meeting; "noted" does not mean "agreed", and each decision is confirmed by email (mine). Do not expect rules to create safety by themselves: O'Donovan and McAuliffe's review of 14 healthcare interventions found mixed results, and education alone struggled to change deeply rooted speaking-up behaviour (O'Donovan and McAuliffe, 2020).
2. Separate the proposal from the person. Say "the current joint design", not the designer's name; each objection states a criterion and a way to test it; the chair takes the first objection and answers with a question, not a defence (mine). The basis is moderate: Edmondson predicts that leaders who respond without defensiveness make teams feel safer (Edmondson, 1999, p. 356), yet the effect of a coaching leader was only close to significant at team level (p = .06) (Edmondson, 1999, p. 368).
3. Measure safety with Edmondson's items, anonymously, in weeks 2 and 6. Three of them: "If you make a mistake on this team, it is often held against you" (reverse scored); "Members of this team are able to bring up problems and tough issues"; "It is safe to take a risk on this team" (Edmondson, 1999, p. 382). Split the scores by side, workshop and buyer (mine). The paper gives no threshold for "safe enough". A wide gap between sides suggests, in my reading, that they are not yet one team, since Edmondson defines the belief as shared by the team (Edmondson, 1999, p. 354).
4. Read the results after six weeks. Friction falling and objections arriving earlier: keep going. Nobody objecting and everybody content: do not celebrate yet, because in Tuckman's model the cohesion stage avoids task conflict to keep harmony (Tuckman, 1965, p. 386). Scores that differ by side: fix steps 1 and 2 before drawing conclusions about anyone's culture.

For the workshop, step 2 turns the engineer's sentence into: "the current joint does not yet meet the durability criterion for one rainy season; who proposes a test?" This suits you if the team works together for months and you chair the meetings; it does not suit you if people meet once, or if power between the sides is so uneven that the weaker side finds it hard to be candid even anonymously (in my view, no source tests this).

Going deeper: what did I read, and what evidence is missing?

I read in full Edmondson, Chua and colleagues, Tuckman 1965, Frazier and colleagues (through copies posted on university servers or repositories) and O'Donovan and McAuliffe (open access). I read only the abstract of Tuckman and Jensen. The points about the input–process–outcome frame come from the library's framework page, not from the original papers by Mathieu and colleagues and Ilgen and colleagues. I could not read Meyer in full.

The main gap: no source above measures task debate and decision quality directly. The library's entry on the two kinds of conflict has no source attached, so I did not use it to conclude that task debate improves decisions. Nor does any source here survey Vietnamese teams, or teams made of two companies in two countries.