← CHI 2026 map

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

Texterial: A Text-as-Material Interaction Paradigm for LLM-Mediated Writing

Jocelyn J Shen , Nicolai Marquardt , Hugo Romat , Ken Hinckley , Nathalie Riche , Fanny Chevalier

DOI PDF Program page Page JSON

Texterial is a strong CHI-style conceptual contribution: it reframes LLM writing around material manipulation rather than prompt entry, and it backs that framing with a formative study plus two evocative probes. The work is most convincing as a design paradigm and less as evidence of downstream effectiveness.

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.

Review Summary

Texterial’s value is in how clearly it challenges the default interaction model for LLM-mediated writing. Rather than treating the prompt box as the primary site of control, it proposes that text itself can be the manipulable substrate, which is a meaningful shift for CHI because it changes both the interaction vocabulary and the mental model users bring to writing with AI. The paper’s conceptual framework is the central contribution: it organizes the space in terms of layers, operations, and material-like mappings, giving designers a way to reason about how AI capabilities might be surfaced through interaction. The two probes, Text as Clay and Text as Plants, are useful because they make the paradigm concrete and show how the metaphor can be instantiated in distinct ways. At the same time, the evidence is appropriately exploratory rather than confirmatory. The formative study is small, and the focus-group evaluation is qualitative, so the paper supports claims about plausibility, interpretability, and design potential more than claims about effectiveness, productivity, or superiority over existing tools. The limitations section matters here: the authors explicitly note the lack of controlled comparisons and the need for longitudinal work, which keeps the contribution grounded. Overall, this reads as a strong honorable-mention-level CHI paper because it offers a memorable conceptual reframing, a coherent framework, and credible early validation, while leaving open the harder questions of deployment and long-term impact.

What Changed

Canon before

Prior CHI work on LLM-mediated writing largely centers on prompt-based interaction, chat-style refinement, and editor-centric assistance; this paper positions text itself as the manipulable medium rather than treating prompting as the primary control surface.

Departure from common sense

The paper’s core move is to reject the default assumption that LLM writing should be steered mainly through linear prompting and chat. Instead, it argues that users can directly manipulate text as a material—through actions like sculpting, stretching, or pruning—to make AI operations legible and to support thinking with the model in a more embodied way.

Actual novelty

The main novelty is a conceptual framework that organizes LLM writing around layers of text meaning and form, the operations that act on those layers, and the mapping from those operations to material-like interactions. The paper also instantiates this framing in two probes, Text as Clay and Text as Plants, to explore how the paradigm can be expressed in interaction design.

Evidence

The paper combines a formative study with 4 participants and a qualitative focus-group study with 10 participants in 4 groups, using two technical probes to explore the paradigm. The evidence supports conceptual and experiential claims about the metaphor and its design implications, but not comparative performance or broad task-level effectiveness.

Limits

Method limits

Evidence comes from a small formative study and a qualitative focus-group evaluation of two probes. The paper does not provide controlled comparisons against conventional editors or AI-assisted tools, and the studies are not designed to establish causal effects or general performance gains.

Deployment limits

The approach is demonstrated through two technical probes rather than a deployable end-user system. Its practical value may depend on whether users can understand and adopt the material metaphor, and whether future implementations can support richer, more robust writing workflows beyond the explored scenarios.

Boundary conditions

The claims are strongest for exploratory writing support and for users who can reason with material metaphors. The paradigm is less validated for routine production writing, long-term use, or contexts where direct prompt-based control and conventional editing affordances remain preferable.

Manual check: false

Position in field

Texterial sits at the intersection of LLM interaction design, writing tools, and tangible/material metaphors. It contributes a design-oriented reframing of AI-assisted writing by shifting attention from prompt engineering to the interactional and conceptual consequences of treating text as a malleable medium.

Abstract

What if text could be sculpted and refined like clay, or cultivated and pruned like a plant? Texterial reimagines text as a material that users can grow, sculpt, and transform. Current generative-AI models enable rich text operations, yet rigid, linear interfaces often mask such capabilities. We explore how the text-as-material metaphor can reveal AI-enabled operations, reshape the writing process, and foster compelling user experiences. A formative study shows that users readily reason with text-as-material, informing a conceptual framework that explains how material metaphors shift mental models and bridge gulfs of envisioning, execution, and evaluation in LLM-mediated writing. We present the design and evaluation of two technical probes: Text as Clay, where users refine text through gestural sculpting, and Text as Plants, where ideas grow serendipitously over time. This work expands the design space of writing tools by treating text as a living, malleable medium.