The early phase of a collaboration is often filled with uncertainty where misunderstandings can harden into structure, shaping what the collaboration can (and cannot) become. Uncertainty is often treated as a risk to be managed or reduced. However, uncertainty is also a starting condition for collaboration and, when handled well, a capacity teams can build. Our data illustrates how teams can be “trained” to work with uncertainty in real time, especially before roles, definitions, and coordination routines stabilize. We examine how structured dialogue prompts make uncertainty visible, discussable, and actionable in the early stages of building large-scale international scientific collaborations.

We analyzed dialogue excerpts from facilitated Toolbox Dialogue sessions with key personnel (PIs, CoPIs, project managers, and graduate students) representing project teams funded by the National Science Foundation’s Accelerating Research through International Network-to-Network Collaborations (AccelNet) program (multiple cohorts 2020 - 2024). These dialogue sessions were organized around questions related to network-of-networks collaboration, communication/coordination, and success conditions. We focused on recurring moments where participants explicitly expressed uncertainty (e.g., “I don’t know,” “we won’t know until we get there”) and traced how those admissions were taken up by the teams. We can map a common progression across teams, beginning how uncertainty was surfaced by the prompt, followed by definitional work (clarifying what key terms mean), followed by inquiry strategies (how to learn what they do not yet know), and finally recognition of systemic/contextual uncertainty tied to international differences.

Across teams, the prompts elicited admissions of uncertainty and, importantly, treated them as legitimate discussion points. Participants often distinguished “known unknowns” from “unknown unknowns,” thereby normalizing the idea that some barriers only appear once collaboration begins, reframing uncertainty as part of responsible planning rather than as incompetence. Once uncertainty was identified and discussed, teams shifted into defining and clarifying these uncertainties, asking what counts as “effective,” what “formal” means across contexts, or what “success” looks like, recognizing that shared action is impossible when core terms carry different meanings across disciplines or countries. From there, uncertainty became a planning task: teams proposed concrete strategies such as explicitly asking partner networks about barriers and motivations, surveying members, clarifying expectations and responsibilities, building feedback loops to update shared understanding as the network evolves. The data also show that uncertainty is often systemic rather than purely informational, linked to language barriers, differing institutional hierarchies, uneven resource control, and contrasting norms for participation in different national research systems. Finally, teams and facilitators repeatedly framed the prompts as a discussion instrument rather than a test, emphasizing that the goal is not consensus or “right answers” but surfacing assumptions and building shared meaning, what we describe as training teams for international collaboration through reflective engagement with uncertainty.

Training teams how to engage with uncertainty during the early stages of a collaboration through reflexive dialogue can provide clarity while building capacity for surfacing and handling uncertainty throughout the project life cycle. This study presents Toolbox dialogue as one method for making uncertainty visible, discussable and actionable that can support collaboration.