← CHI 2026 map

CHI '26 · Honorable mention · full-paper review · confidence medium-high

KNIT: Computational Boundary Objects for Real-Time Convergence in Interdisciplinary Teams

Echo Chuqiao Wan , Carrie Yin , Akira Ito , Ziwei Gao , Jasper Jia , Yuki Taoka , Shigeki Saito , Malak Sadek , Céline Mougenot

KNIT is strongest as a conceptual and design contribution: it reframes AI as a representational mediator that helps teams negotiate meaning, not just automate tasks. The evaluation is credible for a CHI workshop-style paper, but the mechanism claim is broader than the evidence can fully isolate, so the paper reads as influential and well-grounded rather than definitive.


Axes Lens

Rare contribution shape, typical evidence profile. The point here is not a score. It is to show what kind of claim the paper makes, and whether the evidence pattern is unusual or baseline in this 268 -review set.

Contribution shape

Knowledge form
generative knowledge typical · 35/268
Novelty type
framework typical · 59/268
Abstraction level
practice typical · 85/268
Generalization target
organizational context typical · 20/268
Validation mode
mixed methods typical · 136/268

Evidence profile

Evidence strength
moderate typical · 105/268
Claim alignment
strong typical · 231/268
Overclaim risk
medium typical · 210/268

Review Summary

KNIT’s most interesting contribution is the shift in how AI is positioned inside collaborative work. Instead of treating the system as a summarizer, recommender, or quasi-teammate, the paper argues for AI-generated artefacts that function as computational boundary objects: shared representations that make disciplinary differences visible and negotiable. That is a meaningful conceptual move because it connects AI-mediated interaction to a long-standing theory of coordination across boundaries, while also giving designers a concrete mechanism to work with. The paper’s evaluation is reasonably strong for the claim it makes: seven early-stage healthtech teams, 28 participants, and 190 interaction episodes analyzed through Carlile’s 3T framework provide enough empirical grounding to show that the system can support syntactic, semantic, and pragmatic boundary crossing in a real workshop setting. The reported rates are impressive, and the design framing around regenerative plasticity and stakeholder-centered reframing is coherent. At the same time, the evidence does not fully isolate causality. The paper’s strongest claims are interpretive and framework-level, not experimental in the narrow sense. It shows that the system worked in context and that participants could use the artefacts to converge, but it does not prove that representational transformation is the only or primary mechanism across settings. The limitations matter: the deployment is narrow, the teams are all early-stage healthtech, and the system depends on participant input while being unable to validate truth claims or detect strategic framing. So the paper is best read as a well-executed and timely contribution to AI-mediated collaboration, with a novel theoretical lens and a plausible design pattern, rather than as a general solution for interdisciplinary teamwork.

What Changed

Canon before

Prior CHI work on AI-supported collaboration often frames AI as summarizer, recommender, facilitator, or decision aid; boundary-object theory has been used to explain cross-disciplinary coordination, but not typically as a computationally regenerated artifact for live convergence in-session.

Departure from common sense

The paper’s core move is to treat AI not as an autonomous teammate or a passive summarizer, but as a live representational mediator that transforms inputs into shared artefacts for negotiation. That is a less common framing than standard productivity or automation narratives, and it shifts the mechanism from doing the work to reshaping what teams can jointly inspect and contest.

Actual novelty

The paper introduces computational boundary objects as AI-generated artefacts with regenerative plasticity, designed to externalise individual inputs into shared representations that surface semantic overlaps, tensions, and stakeholder reframings. The novelty is not only the label, but the decomposition into boundary-crossing mechanisms and the claim that representational transformation is the primary enabling mechanism.

Evidence

The paper grounds its claims in a workshop evaluation with seven early-stage healthtech teams (28 participants), analyzing 190 interaction episodes through Carlile’s 3T framework. It reports boundary-crossing rates across syntactic, semantic, and pragmatic levels and pairs these results with a design account of how KNIT’s artefacts mediate convergence. The evidence supports a framework-level contribution, though the deployment context is narrow and the mechanism claims are interpretive rather than experimentally isolated.

“ KNIT introduces Computational Boundary Objects : AI-generated artefacts characterised by regenerative plasticity, dynamically reframing team input, surfacing semantic overlaps and tensions, and catalysing negotiated alignmen”

actual novelty · Abstract + Introduction + Design/Evaluation framing · confidence 0.72

“ Rather than positioning AI as either a passive tool or an autonomous teammate, KNIT functioned as a cognitive mediator [ 46 ]”

departure from common sense · Abstract/Introduction + Discussion (6.4) · confidence 0.66

“ While KNIT successfully aligned language across team members, it cannot validate truth claims or detect strategic framing”

limitation · Limitations and Future Work (6.6) · confidence 0.90

“ We evaluated KNIT in workshops with seven early-stage healthtech teams (28 participants), analysing 190 interaction episodes using Carlile’s 3T framework”

validation scope · Abstract + Results (Table 6) · confidence 0.84

Limits

Method limits

The evaluation is workshop-based and interpretive, using 190 interaction episodes from seven teams rather than a controlled comparison that isolates causal effects of specific components. The mechanism claim is supported by qualitative/analytic framing and reported rates, but not by a randomized or ablation-style test of representational transformation versus alternative AI mediation strategies.

Deployment limits

The system was evaluated with seven early-stage UK healthtech teams in workshops, so transfer to other domains, longer-term collaborations, different organizational cultures, or high-stakes settings remains uncertain. The paper also notes that AI outputs depend on participant input and cannot validate truth claims or detect strategic framing.

Boundary conditions

The approach appears most suitable when teams need rapid in-session convergence across disciplinary perspectives and can tolerate AI-generated reframings as negotiable artefacts rather than authoritative outputs. Its value depends on participants’ willingness to externalize inputs, inspect differences, and engage in negotiation around shared representations.

Position in field

This sits at the intersection of CSCW/CHI collaboration support, AI-mediated interaction, and boundary-object theory. Its main contribution is to reframe AI from automation toward representational mediation, offering a conceptual and design vocabulary for AI systems that support interdisciplinary convergence rather than merely summarizing or coordinating work.

Abstract