A Framework for LLM-powered Design Assistants

Swaroop Panda

REVIEW 3 major objections 8 minor 34 references

A Framework for LLM-powered Design Assistants

T0 review · 3 major / 8 minor · reviewed 2026-08-08 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read This paper proposes a framework in which large language models assist designers through three modalities—idea exploration, dialogue with designers, and design evaluation—adaptable to any design process while keeping the designer in control.

desk verdict A coherent but purely conceptual taxonomy of LLM design-assistant roles; the real soft spot is not the missing benchmarks but an adaptability claim with no defined coverage condition. read the letter →

arxiv 2502.07698 v1 pith:VP4XYAFP submitted 2025-02-11 cs.HC cs.CY

Swaroop Panda This is my paper · ORCID

classification cs.HCcs.CY

keywords designassistantslargelanguagemodelsprocessideaexplorationevaluationhuman-AIcollaborationcreativeworkflow

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper proposes a framework for using large language models as design assistants, organized around three modalities: Idea Exploration, Crafting Dialogue with Designers, and Design Evaluation. The central claim is that these modalities are not tied to any single design process; they can be integrated partially or fully into whatever workflow a designer already uses. The paper argues that LLMs' language and data-processing strengths suit them for generating new and historical ideas, checking cultural sensitivity, clarifying ambiguity, and critiquing designs from raw files. It matters because, if the framework holds, designers gain a process-agnostic assistant that augments rather than replaces their judgment. The paper delivers the structure and the arguments for it, not experimental validation.

What carries the argument

The central object is the three-modality framework, visualized in a figure adapted from a model of design processes [28]. Its mechanism is a mapping from design-process checkpoints to LLM capabilities: each assistance opportunity is classified into Idea Exploration, Dialogue with Designers, or Design Evaluation, and each modality is broken into concrete tasks the LLM is expected to perform. The framework is deliberately process-agnostic, so it works by attaching itself to existing workflows rather than prescribing a new one, with the designer always retaining final authority.

What would settle it

Run a controlled benchmark where the same design brief goes through all three modalities under two different design methodologies; if the LLM outputs are judged unusable by designers or no better than a non-LLM baseline on tasks such as dark-pattern detection and SVG critique, the framework's core assumption of reliable, process-independent assistance is falsified.

Watch

Extended reading notes

Core claim

The paper's central claim is that the design process can be read as a series of checkpoints at which an LLM can help without taking control, and that all such help points fall into three modalities. Idea Exploration covers generating new ideas, recovering old ones, assessing cultural sensitivities, and scanning competition. Crafting Dialogue with Designers covers clarifying ambiguities, working through ethical dilemmas, and overcoming designer's block. Design Evaluation covers critiquing raw design files such as SVG and CSV, providing a full critique of design choices, comparing designs against existing products, and detecting dark patterns. The author's claim is that these three modalities are comprehensive and adaptable, so the framework remains useful across different design methodologies and organizational contexts.

Load-bearing premise

The framework assumes that LLMs can actually perform the listed tasks—generating novel and historically grounded ideas, reading cultural cues, clarifying ambiguity, parsing SVG and CSV files, spotting dark patterns, and comparing designs—reliably enough to help, and that this reliability transfers across arbitrary design processes.

Editorial extensions

If this is right

Design tools can embed LLM assistance at many points without forcing designers into a fixed pipeline, since the checkpoints are process-agnostic.
Each modality gives a separate target for evaluation: idea quality, dialogue helpfulness, and critique accuracy can be measured independently.
The framework predicts that the most useful LLM design assistance is organized around these three roles rather than around a single text-generation task.
Because the framework keeps the designer in control, it implies assistant designs that advise and explain rather than automate final decisions.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The framework is a taxonomy rather than a tested system; its real payoff depends on whether each listed capability survives empirical benchmarking with human designers.
The same three modalities could map onto other creative domains and onto non-LLM generative models, so the structure may be broader than the design examples given.
The emphasis on designer control implies an unstated requirement for transparency and override mechanisms; specifying and testing those would be a natural extension.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Van, quick read of Panda (arXiv:2502.07698). Bottom line: this is a tidy organizational paper, not a research result. The three modalities—Idea Exploration, Dialogue, Evaluation—and their sub-activities are all present in the cited literature (Shaer et al. on ideation, Luther et al. on critique, Gray et al. on dark patterns, Xu and Wall on SVG parsing). The author recombines them into a clean taxonomy with a sensible Figure 1 adapted from Takeda et al. As a synthesis, it is accurate and well-cited; a newcomer to the area could use it as a map. Credit where due: the writing is clear, the citations are relevant, and the framework does not overclaim what individual LLM capabilities are known to do. That is real value, but modest.

The soft spot the reader flagged is real but secondary: no user study, benchmark, or error analysis supports the claim that LLMs can do these tasks reliably. That alone would justify a conditional verdict. The stress-test note lands harder, and I think it is the actual load-bearing issue: the central claim that the framework is 'adaptable across various processes' is unfalsifiable as stated. Coverage is never defined. There is no characterization of what it means for a design process to instantiate a modality or checkpoint, so no conceivable process could fail to fit. Section 3 says the framework is 'devoid of any presupposition regarding a specific design process' and Section 5 calls it a 'dynamic and adaptive guideline.' But a guideline compatible with every process is indistinguishable from one compatible with none unless the paper specifies testable instantiation conditions. Even with perfect LLMs, the adaptability thesis would remain a tautology. That is the core problem, and the reader’s reliability concern, while valid, is secondary.

I also note the definitional looseness: 'Crafting Dialogue with Designers' lists clarifying ambiguities, ethical dilemmas, and designer’s block—activities that are not clearly a single modality, and the ethical-dilemma item leans on a citation about group ideation rather than ethics scholarship. Minor, but symptomatic of a paper where categories are assembled rather than derived.

Who is this for? Graduate students or practitioners wanting a quick survey of where LLMs might plug into design workflows. It is a competent literature-organization piece, not a contribution that needs rigorous refereeing in its current form. My recommendation: desk reject, but invite the author to resubmit a revised version that either (a) grounds the adaptability claim by mapping the three modalities onto at least two concrete design processes with instancing criteria, or (b) reframes the paper honestly as a survey, which would make it publishable in a workshop venue. If the author does either, it deserves serious referee time then; as it stands, the unfalsifiable adaptability claim makes the central thesis vacuous.

Referee Report

3 major / 8 minor

Summary. This paper proposes a framework for positioning large language models as design assistants, organized around three modalities: Idea Exploration, Dialogue with Designers, and Design Evaluation. Each modality is decomposed into sub-activities such as generating new ideas, analyzing cultural sensitivities, clarifying ambiguities, parsing SVG and CSV design files, performing comparative evaluation, and detecting dark patterns. The paper emphasizes that the framework is not tied to any particular design process and can be partially or fully integrated into diverse design methodologies. The contribution is conceptual: Sections 3 and 4 describe the anatomy and the modalities, and Section 5 characterizes the framework as a dynamic and adaptive guideline. No experiments, user studies, benchmarks, or worked instantiations are reported.

Significance. The paper addresses a timely question: how to place LLMs systematically within design workflows rather than deploying them ad hoc. The modality decomposition is reasonable, and the paper grounds its motivation in canonical design literature (Simon, Cross, Dorst and Cross, Lawson) while citing recent LLM-HCI research. It is also a strength that the framework explicitly scopes the assistant as augmenting rather than replacing the designer, keeping the designer in control of final decisions. The framework is internally coherent as a taxonomy. The limiting factor is that the central claims are argued by plausibility rather than demonstration: the adaptability claim is unfalsifiable as stated, and the capability claims in Section 4 are asserted without evidence. If the generality claim were sharpened and at least one concrete instantiation were provided, the paper could serve as a useful organizing device; in its present form it is best read as a checklist of opportunities rather than a validated framework.

major comments (3)

[Section 3 (also Abstract and Section 5)] The paper's headline claim that the framework is 'adaptable across various processes' is unfalsifiable as stated. Section 3 says the framework is 'structured around various checkpoints' and 'intentionally designed to be adaptable, devoid of any presupposition regarding a specific design process,' but the paper never defines what it means for a design process to contain a checkpoint or to instantiate a modality, and it never specifies coverage conditions that would let a reader identify a design process for which the framework fails. Since almost any sequence of design activities can be retrospectively labeled as 'Idea Exploration,' 'Dialogue with Designers,' or 'Design Evaluation,' the abstract's claim that the framework 'is not confined to a singular design process' risks being a tautology rather than a substantive property. Section 5's self-characterization as a 'dynamic and adaptive guideline' reinforces this concern: a guideline without instantiation criteria is not a framework with a testable generality claim. The figure provenance compounds the issue: Figure 1 is adapted from Takeda et al.'s design-process model [28], a specific process model, yet the text claims no presupposition of a specific design process; the paper should either explain how the figure generalizes or acknowledge the dependence on a particular model. To make the central claim testable, the authors should define modality instantiation (for example, as a mapping from a process's stages or activities to the three modalities), specify the class of design processes to which the claim applies, and demonstrate the mapping on at least one concrete process.
[Sections 4.1–4.3] The framework's utility rests on the claim that LLMs can actually perform the listed tasks, but this is asserted rather than supported. Section 4 is written in the indicative: LLMs 'possess the capability to comprehend and generate language across multiple cultural contexts' (Section 4.1.2), 'can conduct a comprehensive critique of the design in its original raw format' from SVG/CSV data (Section 4.3.1), and 'can be finetuned to recognize and classify' dark patterns (Section 4.3.4). The cited references are largely exploratory or adjacent: [1] is a survey of how to measure culture in LLMs, [32] is an exploratory study of SVG visual-analytic tasks, and [13] documents dark patterns rather than LLM-based detection. No experiments, benchmarks, user studies, or error analyses are reported, and the paper does not address known failure modes such as hallucination or bias, despite Section 5 acknowledging that LLM limitations require 'ongoing reassessment.' If the paper is intended as a conceptual framework, the capability statements should be explicitly reframed as hypotheses or opportunities; if it is intended to support the stronger claim that LLM-powered design assistants are usable in practice, evidence is required. In either case, the overstatement in Section 4.3.3 that the approach ensures 'a more objective and reliable assessment' by mitigating 'human bias and error' should be removed or qualified, since the cited work does not establish this and model bias is a documented concern.
[Section 5 (Discussion)] The paper provides no criteria for applying the framework in practice. Section 5 states that the framework is a 'dynamic and adaptive guideline' requiring 'ongoing reassessment' as LLMs evolve, but it does not say how a designer or researcher should decide when to include a modality, how to evaluate whether the LLM's contribution at a checkpoint is successful, or how the framework should be revised in response to empirical insights. Without an adaptation protocol or success criteria, the framework cannot be deployed or tested by others; adding a short operationalization (for example, integration questions, an evaluation rubric per modality, or a revision procedure) would make the proposal actionable.

minor comments (8)

[Section 4.3.1] The heading 'Engaging directly with DesignLLMs' contains a missing space and introduces the undefined term 'DesignLLMs'; the subsection actually describes data parsing and design critique rather than direct engagement with a design artifact.
[Section 4.1.1] There is a stray space before the comma in 'tailored prompts , LLMs'.
[Abstract and Section 4.2] The second modality is named 'Dialogue with Designers' in the Abstract and Section 1 but 'Crafting Dialogue with Designers' in Section 4.2; the naming should be unified.
[Figure 1] Figure 1 is never referenced or discussed in the body text, and the caption does not explain how the three modalities relate to the diagram adapted from Takeda et al. [28].
[Sections 4.1.3–4.1.4] Some citation placements are awkward or missing: 'LLMs can analyze [27] archival texts' interrupts the sentence, and the claims about collaborative platforms and crowdsourcing in Section 4.1.4 have no supporting citation.
[Section 5] The paper ends abruptly after Section 5; a short conclusions paragraph restating contributions and limitations would help the reader.
[Section 2.2] Several broad capability claims in Section 2.2 (for example, that LLMs 'enhance collaborative design environments') are made without citations; supporting references or hedging would be appropriate.
[Section 4] The opening sentence of Section 4 uses a semicolon where a colon is expected before the list of modalities.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the paper is a qualitative framework proposal with no derivation chain to reduce to its own inputs.

full rationale

The paper is a qualitative position and synthesis piece. It proposes three modalities (Idea Exploration, Crafting Dialogue with Designers, and Design Evaluation) and enumerates possible LLM-assisted activities, each supported by external citations. There is no derivation chain: no equations, no fitted parameters, no quantity predicted from fitted inputs, no uniqueness theorem, and no self-citations at all. The closest potential concern is the adaptability claim in the abstract and Section 3 ("our framework is not confined to a singular design process but is adaptable across various processes") and the Discussion's description of the framework as "a dynamic and adaptive guideline." That claim is under-specified and arguably unfalsifiable, but unfalsifiability is a precision or validity risk, not circularity: the claim is asserted rather than derived from a definition that presupposes it, and no result is forced by construction. The framework does not reduce to its inputs by definition, and it makes no predictions that could be statistically forced. Therefore, despite the absence of empirical validation, there is no significant circularity and the score is 0.

Assumptions & free parameters 0 free parameters · 2 assumptions · 0 invented entities

The paper introduces no new entities and fits no free parameters. It relies on two domain assumptions: that LLMs possess the cited capabilities and that a generic three-modality framework transfers across design processes. Both are unvalidated in this preprint.

assumptions (2)

domain assumption LLMs can reliably perform the listed design-assistance tasks, including idea generation, cultural sensitivity analysis, dialogue repair, design critique, and dark-pattern detection.
Introduced in Sections 4.1, 4.2, and 4.3 without empirical tests, benchmarks, or user studies; the paper takes these capabilities as given from cited work.
domain assumption A single framework with these three modalities is adaptable to arbitrary design processes.
Asserted in Sections 3 and 5 without comparing different design processes or testing cross-process transfer.

how reviews work

0 comments

Cite this review

Pith. "Pith review of A Framework for LLM-powered Design Assistants." pith.science (2026). https://pith.science/paper/VP4XYAFP

@misc{pith2026250207698,
  author       = {Pith},
  title        = {Pith review of: A Framework for LLM-powered Design Assistants},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VP4XYAFP}},
  note         = {Machine review of arXiv:2502.07698}
}

read the original abstract

Design assistants are frameworks, tools or applications intended to facilitate both the creative and technical facets of design processes. Large language models (LLMs) are AI systems engineered to analyze and produce text resembling human language, leveraging extensive datasets. This study introduces a framework wherein LLMs are employed as Design Assistants, focusing on three key modalities within the Design Process: Idea Exploration, Dialogue with Designers, and Design Evaluation. Importantly, our framework is not confined to a singular design process but is adaptable across various processes.

Figures

Figures reproduced from arXiv: 2502.07698 by the authors.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

34 extracted references · 26 canonical work pages

[28]

Takeda, P

H. Takeda, P. Veerkamp, and H. Yoshikawa. Modeling design process. AI magazine, 11(4):37–37, 1990

work page 1990
[1]

M. F. Adilazuarda, S. Mukherjee, P. Lavania, S. Singh, A. Dwivedi, A. F. Aji, J. O’Neill, A. Modi, and M. Choudhury. Towards measuring and modeling” culture” in llms: A survey. arXiv preprint arXiv:2403.15412 , 2024

arXiv 2024
[32]

Exploring the Capability of LLMs in Performing Low-Level Visual Analytic Tasks on SVG Data Visualizations

Z. Xu and E. Wall. Exploring the capability of llms in performing low- level visual analytic tasks on svg data visualizations. arXiv preprint arXiv:2404.19097, 2024

work page Pith review arXiv 2024
[13]

C. M. Gray, Y. Kou, B. Battles, J. Hoggatt, and A. L. Toombs. The dark (patterns) side of ux design. In Proceedings of the 2018 CHI conference on human factors in computing systems , pages 1–14, 2018

work page 2018
[2]

Alabood, Z

L. Alabood, Z. Aminolroaya, D. Yim, O. Addam, and F. Maurer. A sys- tematic literature review of the design critique method. Information and Software Technology, 153:107081, 2023

work page 2023
[3]

Aubin Le Qu´ er´ e, H

M. Aubin Le Qu´ er´ e, H. Schroeder, C. Randazzo, J. Gao, Z. Epstein, S. T. Perrault, D. Mimno, L. Barkhuus, and H. Li. Llms as research tools: Applications and evaluations in hci data work. In Extended Abstracts of the CHI Conference on Human Factors in Computing Systems , pages 1–7, 2024

work page 2024
[4]

Barua, C

A. Barua, C. Widmer, and P. Hitzler. Concept induction using llms: a user experiment for assessment. arXiv preprint arXiv:2404.11875 , 2024

arXiv 2024
[5]

Bj¨ orkman

H. Bj¨ orkman. Design dialogue groups as a source of innovation: factors behind group creativity. Creativity and Innovation Management , 13(2): 97–108, 2004

work page 2004

Show all 34 references

[6]

Z. G. Cai, D. A. Haslett, X. Duan, S. Wang, and M. J. Pickering. Does chatgpt resemble humans in language use? 2023

2023
[7]

C ¸ elen, G

A. C ¸ elen, G. Han, K. Schindler, L. Van Gool, I. Armeni, A. Obukhov, and X. Wang. I-design: Personalized llm interior designer. arXiv preprint arXiv:2404.02838, 2024

2024
[8]

R. Chew, J. Bollenbacher, M. Wenger, J. Speer, and A. Kim. Llm-assisted content analysis: Using large language models to support deductive coding. arXiv preprint arXiv:2306.14924 , 2023

2023 arXiv
[9]

N. Cross. Designerly ways of knowing. Design studies, 3(4):221–227, 1982

1982
[10]

Dorst and N

K. Dorst and N. Cross. Creativity in the design process: co-evolution of problem–solution. Design studies , 22(5):425–437, 2001

2001
[11]

P. Duan, J. Warner, and B. Hartmann. Towards generating ui design feed- back with llms. InAdjunct Proceedings of the 36th Annual ACM Symposium on User Interface Software and Technology , pages 1–3, 2023

2023
[12]

Gramopadhye, S

O. Gramopadhye, S. S. Nachane, P. Chanda, G. Ramakrishnan, K. S. Jad- hav, Y. Nandwani, D. Raghu, and S. Joshi. Few shot chain-of-thought driven reasoning to prompt llms for open ended medical question answer- ing. arXiv preprint arXiv:2403.04890 , 2024. 8

2024 arXiv
[14]

A. Huet, F. Segonds, R. Pinquie, P. Veron, J. Guegan, and A. Mallet. Context-aware cognitive design assistant: Implementation and study of de- sign rules recommendations. Advanced Engineering Informatics, 50:101419, 2021

2021
[15]

H. Jin, Y. Zhang, D. Meng, J. Wang, and J. Tan. A comprehensive survey on process-oriented automatic text summarization with exploration of llm- based methods. arXiv preprint arXiv:2403.02901 , 2024

2024
[16]

S.-G. Kim, S. M. Yoon, M. Yang, J. Choi, H. Akay, and E. Burnell. Ai for design: Virtual design assistant. CIRP Annals, 68(1):141–144, 2019

2019
[17]

Koyama and M

Y. Koyama and M. Goto. Bo as assistant: Using bayesian optimization for asynchronously generating design suggestions. In Proceedings of the 35th Annual ACM Symposium on User Interface Software and Technology , pages 1–14, 2022

2022
[18]

B. Lawson. How designers think: The process demystified. Translated by H. Nadimi, Tehran, beheshti university , 2005

2005
[19]

C. Lee, S. Kim, D. Han, H. Yang, Y.-W. Park, B. C. Kwon, and S. Ko. Guicomp: A gui design assistant with real-time, multi-faceted feedback. In Proceedings of the 2020 CHI conference on human factors in computing systems, pages 1–13, 2020

2020
[20]

J. Li, T. Tang, W. X. Zhao, J.-Y. Nie, and J.-R. Wen. Pre-trained language models for text generation: A survey. ACM Computing Surveys , 56(9):1– 39, 2024

2024
[21]

Luther, J.-L

K. Luther, J.-L. Tolentino, W. Wu, A. Pavel, B. P. Bailey, M. Agrawala, B. Hartmann, and S. P. Dow. Structuring, aggregating, and evaluating crowdsourced design critique. In Proceedings of the 18th ACM conference on computer supported cooperative work & social computing, pages...

2015
[22]

Chatgpt: Optimizing language models for dialogue, 2023

OpenAI. Chatgpt: Optimizing language models for dialogue, 2023. URL https://openai.com/chatgpt. Accessed: 2025-02-11

2023
[23]

M. Rose. Writer’s block: The cognitive dimension . SIU Press, 2009

2009
[24]

Shaer, A

O. Shaer, A. Cooper, O. Mokryn, A. L. Kun, and H. Ben Shoshan. Ai- augmented brainwriting: Investigating the use of llms in group ideation. In Proceedings of the CHI Conference on Human Factors in Computing Systems, pages 1–17, 2024

2024
[25]

H. A. Simon. The sciences of the artificial mit press. Cambridge, Ma, 1969. 9

1969
[26]

Stella, C

F. Stella, C. Della Santina, and J. Hughes. How can llms transform the robotic design process? Nature Machine Intelligence , 5(6):561–564, 2023

2023
[27]

R. H. Tai, L. R. Bentley, X. Xia, J. M. Sitt, S. C. Fankhauser, A. M. Chicas-Mosier, and B. G. Monteith. An examination of the use of large language models to aid analysis of textual data. International Journal of Qualitative Methods, 23:16094069241231168, 2024

2024
[29]

V´ azquez

P.-P. V´ azquez. Are llms ready for visualization? In2024 IEEE 17th Pacific Visualization Conference (PacificVis) , pages 343–352. IEEE, 2024

2024
[30]

Whitfield and M

S. Whitfield and M. A. Hofmann. Elicit: Ai literature review research assistant. Public Services Quarterly , 19(3):201–207, 2023

2023
[31]

X. Xu, J. Yin, C. Gu, J. Mar, S. Zhang, J. L. E, and S. P. Dow. Jamplate: Exploring llm-enhanced templates for idea reflection. In Proceedings of the 29th International Conference on Intelligent User Interfaces , pages 907– 921, 2024

2024
[33]

Zhang, C

H. Zhang, C. Wu, J. Xie, Y. Lyu, J. Cai, and J. M. Carroll. Redefining qualitative analysis in the ai era: Utilizing chatgpt for efficient thematic analysis. arXiv preprint arXiv:2309.10771 , 2023

2023 arXiv
[34]

W. X. Zhao, K. Zhou, J. Li, T. Tang, X. Wang, Y. Hou, Y. Min, B. Zhang, J. Zhang, Z. Dong, et al. A survey of large language models. arXiv preprint arXiv:2303.18223, 2023. 10

2023 arXiv

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.

[1] [28]

Takeda, P

H. Takeda, P. Veerkamp, and H. Yoshikawa. Modeling design process. AI magazine, 11(4):37–37, 1990

work page 1990

[2] [1]

M. F. Adilazuarda, S. Mukherjee, P. Lavania, S. Singh, A. Dwivedi, A. F. Aji, J. O’Neill, A. Modi, and M. Choudhury. Towards measuring and modeling” culture” in llms: A survey. arXiv preprint arXiv:2403.15412 , 2024

arXiv 2024

[3] [32]

Exploring the Capability of LLMs in Performing Low-Level Visual Analytic Tasks on SVG Data Visualizations

Z. Xu and E. Wall. Exploring the capability of llms in performing low- level visual analytic tasks on svg data visualizations. arXiv preprint arXiv:2404.19097, 2024

work page Pith review arXiv 2024

[4] [13]

C. M. Gray, Y. Kou, B. Battles, J. Hoggatt, and A. L. Toombs. The dark (patterns) side of ux design. In Proceedings of the 2018 CHI conference on human factors in computing systems , pages 1–14, 2018

work page 2018

[5] [2]

Alabood, Z

L. Alabood, Z. Aminolroaya, D. Yim, O. Addam, and F. Maurer. A sys- tematic literature review of the design critique method. Information and Software Technology, 153:107081, 2023

work page 2023

[6] [3]

Aubin Le Qu´ er´ e, H

M. Aubin Le Qu´ er´ e, H. Schroeder, C. Randazzo, J. Gao, Z. Epstein, S. T. Perrault, D. Mimno, L. Barkhuus, and H. Li. Llms as research tools: Applications and evaluations in hci data work. In Extended Abstracts of the CHI Conference on Human Factors in Computing Systems , pages 1–7, 2024

work page 2024

[7] [4]

Barua, C

A. Barua, C. Widmer, and P. Hitzler. Concept induction using llms: a user experiment for assessment. arXiv preprint arXiv:2404.11875 , 2024

arXiv 2024

[8] [5]

Bj¨ orkman

H. Bj¨ orkman. Design dialogue groups as a source of innovation: factors behind group creativity. Creativity and Innovation Management , 13(2): 97–108, 2004

work page 2004