Script&Shift: A Layered Interface Paradigm for Integrating Content Development and Rhetorical Strategy with LLM Writing Assistants
Momin Siddiqui, Roy Pea, Hari Subramonyam · 2025
Read Script&Shift: A Layered Interface Paradigm for Integrating Content Development and Rhetorical Strategy with LLM Writing Assistants by Momin Siddiqui, Roy Pea, Hari Subramonyam (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2502-10638.txt.
Introduction
Momin Siddiqui, Roy Pea, and Hari Subramonyam present a layered writing interface that pairs human writers with language models, in a paper from twenty twenty-five. Good writing reshapes ideas through planning, drafting, and review. AI help can speed that work, but it can also break the writer's flow, above all on tasks like moving text across sections.
Their design sorts help into layers matched to what the writer means to do. One mode grows content, while another reshapes structure or retunes text for a new audience. The split aims to keep the writer in charge and invite playful trial and revision. Tests show writers could fold model help into their work smoothly and creatively, without losing their thread.
Catalog record
| CREATOR | Momin Siddiqui, Roy Pea, Hari Subramonyam |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 76 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2502-10638
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Script&Shift: A Layered Interface Paradigm for Integrating Content Development and Rhetorical Strategy with LLM Writing Assistants Momin Siddiqui, Roy Pea, Hari Subramonyam Abstract . Good writing is a dynamic process of knowledge transformation, where writers refine and evolve ideas through planning, translating, and reviewing. Generative AI-powered writing tools can enhance this process but may also disrupt the natural flow of writing, such as when using LLMs for complex tasks like restructuring content across different sections or creating smooth transitions. We introduce Script&Shift, a layered interface paradigm designed to minimize these disruptions by aligning writing intents with LLM capabilities to support diverse content development and rhetorical strategies. By bridging envisioning, semantic, and articulatory distances, Script&Shift’s interactions allow writers to leverage LLMs for various content development tasks ( scripting ) and experiment with diverse organization strategies while tailoring their writing for different audiences ( shifting ). This approach preserves creative control while encouraging divergent and iterative writing. Our evaluation shows that Script&Shift enables writers to creatively and efficiently incorporate LLMs while preserving a natural flow of composition. Conference: CHI Conference on Human Factors in Computing Systems; April 26-May 1, 2025; Yokohama, Japan CHI Conference on Human Factors in Computing Systems (CHI ’25), April 26-May 1, 2025, Yokohama, Japan DOI: 10.1145/3706598.3714119 ISBN: 979-8-4007-1394-1/25/04 CCS: Human-centered computing Systems and tools for interaction design CCS: Human-centered computing Empirical studies in interaction design Figure 1 . Script&Shift is an AI-assisted writing interface that empowers writers to query LLMs using familiar visual design elements. (A) Scripting enables users to query several specialized Writer’s Friends inline. (B) Shifting allows writers to reorganize layers and combine them to see what the final document looks like. 1. Introduction Writing is rarely a linear process. Instead, it involves a complex interplay of cognitive activities spanning various levels of abstractions and perspectives ( Kellogg, 1999 ; Pea and Kurland, 1987a ) . Writers must juggle fine-grained details — such as word choice and sentence structure — while simultaneously managing broader concerns such as overall organization and rhetorical strategy ( Hayes, 2012 ; Hayes and Flower, 1980 ) . For instance, imagine writing the introduction to this CHI paper. The author might start by crafting a compelling opening sentence (e.g., “Large Language Models (LLMs) are effective writing collaborators.” ), only to pause and reconsider how this framing aligns with the paper’s overall contribution. They might zoom out to then enumerate what the key contributions are before diving back in to flesh out a particular argument. As they write, they must anticipate potential questions and critiques from readers, adjusting their narrative to address each of these imagined interlocutors. This dynamic interplay between knowledge creation and rhetorical problem-solving is what distinguishes good writing as a process of knowledge transformation rather than mere knowledge telling ( Bereiter and Scardamalia, 1987 ) . However, the quality of writing tools can make or break this dynamic process of composition, largely depending on the affordances they provide – or fail to provide ( Kozma, 1991 ; Kellogg, 1999 ) . Tools such as pen and paper afford a flexible tactile medium for writing, but quickly become limiting when writers need to make extensive revisions or rearrange large blocks of text. Word processors, such as Microsoft Word, allow easy editing and formatting, but the page-based metaphor makes it challenging to escape the rigid top-to-bottom organization. For instance, in the CHI paper example, the author may want to try multiple ways of presenting the argument, such as structuring it around the design process, or components of a theoretical framework. Similarly, they may want to explore alternative arguments to benefits of LLMs in writing. However, such a process can be labor intensive and require duplicating sections, manually copying content, or creating entirely new documents just to assess how these alternative structures and arguments might develop. As a third alternative, platforms such as Overleaf allow writers to manage large sections of papers more efficiently through modular organization and referential code-like syntax. On the other hand, the precise formatting requirements can hinder exploratory writing. Ultimately, writers are left with tools that offer affordances for certain tasks, but which hinder accomplishing others. As a result, writers often avoid engaging in iterative, exploratory, and divergent writing, even though exploratory writing ( Elbow, 1998 ; Bean and Melzer, 2021 ) can produce good writing ( Bereiter and Scardamalia, 1987 ) . As evidenced by rapidly expanding landscape of LLM writing tools ( Lee et al., 2024 ) , LLMs hold the potential to alleviate many of the process-related challenges, but interface issues largely remain. At the process level, writers can prompt LLMs to generate alternative phrasings, improve sentence clarity, or provide stylistic variations, enabling more efficient iteration on the small details without becoming bogged down in manual rewriting ( Reza et al., 2023 ) . For instance, while drafting a section of this CHI paper, the author could request the model to suggest multiple ways to introduce a complex argument, allowing for immediate exploration of tone and clarity. Or they might use LLMs to receive feedback from different perspectives ( Benharrak et al., 2024 ) as well as structure suggestions for their text. Despite these benefits, the exceptional potential of LLMs is often constrained by the traditional, linear paradigms of existing writing tools or multi-turn conversational interfaces. These LLM environments don’t naturally support the fluidity required for expansive writing. For instance, prompting LLMs across content levels remains a challenge due to these rigid document structures. A writer may wish to ask an LLM to refine a sentence within a section while simultaneously considering how that change affects the cohesion in a different section of the paper. To achieve this aim, they would need to prompt the LLM twice: first, referencing the specific sentence or paragraph for fine-grained revision, and then referencing the broader section or entire document to assess how the changes influence overall coherence. This process can be challenging because current writing tools often do not support seamless cross-referencing between micro and macro levels. Writers must manually manage the integration of these prompts, which disrupts the flow ballistics of the writing process. Therefore, the motivating question for our work is: How can writing tools be designed to better support seamless, iterative prompting across both micro and macro levels without disrupting the writer’s workflow and associated cognitive flow? We introduce Script&Shift, a novel layered interface paradigm for LLM-augmented writing. Script&Shift is designed to address the limitations of traditional writing tools by offering a layered, zoomable workspace that allows writers to seamlessly interleave content development and structural organization. Our layered paradigm supports flexible prompting across both micro and macro levels, allowing users to issue complex queries to LLMs for content generation, restructuring, and refinement. As shown in Figure 1 Script&Shift reduces the gulf of envisioning ( Subramonyam et al., 2024 ) and execution ( Hutchins et al., 1985 ) by offering intuitive interactions such as inline friends, contextual toolbars, intelligent structuring, and other layer-based user interactions such as stacking and folding. These features make it easier for writers to issue complex prompts to LLMs and receive targeted feedback, streamlining the management of complex writing tasks. Results from our usability study and comparative evaluation show that Script&Shift supports expressive