Pith. sign in

REVIEW 3 minor 30 references

An open-source self-hosted platform generates generative AI stimuli for conjoint analysis surveys.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.3

2026-06-27 06:01 UTC pith:FAFL5FN2

load-bearing objection This paper ships a working open-source platform for conjoint analysis that adds LLM text and image stimuli plus full prompt exports, which is the real contribution.

arxiv 2606.12972 v1 pith:FAFL5FN2 submitted 2026-06-11 cs.HC

From Prompts to Preferences: An Open-Source Platform for Generative AI-Enhanced Conjoint Analysis

Philipp Brauner This is my paper
classification cs.HC
keywords conjoint analysisgenerative AIopen source platformpreference elicitationhuman computer interactionvisual stimuliLLM generated content
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper introduces a web application that allows researchers to design, run, and analyze conjoint surveys using AI to create textual descriptions and visual images based on attribute profiles. This addresses the barrier of expensive commercial tools by being free and self-hosted, with features like a setup wizard and AI suggestions for attributes. A proof-of-concept study on preferences for care robots demonstrates the use of AI-generated visuals. The work stresses that researchers must still provide the theoretical foundation while the platform handles the technical generation and data handling.

Core claim

The platform uses a base prompt parameterized by the conjoint profile, with optional level annotations, to generate integrated textual and visual stimuli via LLMs and text-to-image models, while providing structured tools and a full export bundle for transparency, as validated in a study with 55 participants on ambient assisted living robots.

What carries the argument

The researcher-defined base prompt parameterized with the conjoint profile and LLM-facing level annotations, which enable generation of non-tabular stimuli while preserving attribute levels.

Load-bearing premise

AI-generated stimuli can be created from conjoint profiles in a way that accurately reflects the specified attribute levels without systematic biases or unwanted artifacts.

What would settle it

An experiment that measures the same preferences using both the platform's AI stimuli and standard tabular formats and finds significant, consistent differences not explained by the intended attributes.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Researchers gain access to end-to-end conjoint infrastructure without commercial costs.
  • Conjoint studies can incorporate immersive textual scenarios and visuals generated from profiles.
  • AI-assisted attribute suggestion and live analysis reduce setup time for new users.
  • Export of all stimuli, prompts, and data supports reproducibility and transparency.
  • Theoretical design responsibility remains with the researcher even as generation is automated.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Future studies could test whether AI stimuli lead to different preference measurements than traditional formats.
  • This platform might facilitate cross-disciplinary use in political science or healthcare by lowering entry barriers.
  • Extensions could include automated analysis features or integration with other survey tools.
  • Validation experiments comparing AI outputs to human-designed stimuli would strengthen claims about bias-free generation.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

0 major / 3 minor

Summary. The manuscript presents an open-source, self-hosted web application for designing, deploying, and analyzing conjoint surveys. It extends standard tabular stimuli by using generative AI (LLMs for textual scenarios and text-to-image models for visuals), with researcher-defined base prompts parameterized by conjoint profiles and optional level annotations. Additional features include a structured setup wizard, AI-assisted attribute suggestion, live data analysis, and a full export bundle containing stimuli, prompts, and response data to support reproducibility. The platform is illustrated via a proof-of-concept study (N=55) on care-robot preferences for ambient assisted living using AI-generated visual stimuli. The authors stress that theoretical grounding and bias checking remain the researcher's responsibility rather than a property of the generative pipeline.

Significance. If the described functionality holds, the contribution meaningfully lowers technical barriers for researchers in HCI, marketing, political science, and healthcare who lack access to commercial conjoint platforms. The open-source release, self-hosting option, and emphasis on exportable artifacts for transparency directly address reproducibility concerns common in preference-elicitation studies. The balanced discussion of AI assistance—explicitly disclaiming automatic validity of generated stimuli—avoids overclaiming while expanding the methodological toolkit for integrated textual-visual designs.

minor comments (3)
  1. [Abstract] Abstract: the proof-of-concept study is described only at a high level (N=55, care-robot domain); adding one sentence on the number of attributes/levels or the analysis method used would give readers an immediate sense of the demonstration's scope without requiring them to reach the methods section.
  2. [Platform description] The manuscript would benefit from an explicit statement (perhaps in the platform-architecture section) of the exact output formats included in the 'full export bundle' (e.g., CSV, JSON, image files, prompt logs) to clarify reproducibility claims.
  3. [Introduction or Related Work] A short table or bullet list comparing the new platform's feature set against the two or three most common existing open-source conjoint tools would help readers quickly assess the incremental contribution.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for their positive and detailed summary of the manuscript, their recognition of the platform's potential to lower barriers for researchers, and their recommendation to accept. We have no major comments to address.

Circularity Check

0 steps flagged

No significant circularity identified

full rationale

The manuscript is a software platform description paper with no equations, derivations, fitted parameters, or predictive claims. The central contribution is the release of an open-source tool whose architecture and features are presented directly; the N=55 study is labeled a proof-of-concept only, and the text explicitly states that theoretical grounding and bias validation remain the researcher's responsibility rather than asserting that the generative pipeline is measurement-valid by construction. No self-citations, ansatzes, or uniqueness theorems are invoked as load-bearing premises. The derivation chain is therefore self-contained and contains no reductions of outputs to inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

This is a software platform and tool-description paper rather than a theoretical or empirical derivation. No free parameters, axioms, or invented entities are required for the central claim of platform existence and features.

pith-pipeline@v0.9.1-grok · 5752 in / 1212 out tokens · 30896 ms · 2026-06-27T06:01:45.259582+00:00 · methodology

0 comments
read the original abstract

Conjoint analysis is a widely used preference measurement method in marketing research, political science, healthcare, and human-computer interaction. Despite broad adoption, researchers without access to commercial platforms face significant barriers, as existing tools are either expensive or lack end-to-end survey infrastructure. This paper presents an open-source, self-hosted web application for designing, deploying, and analysing conjoint surveys. Beyond conventional tabular stimuli, the platform uses generative AI to produce integrated stimuli formats: textual scenario descriptions generated by a large language model, and visual stimuli by a text-to-image model. A researcher-defined base prompt is parameterised with the conjoint profile, and optional LLM-facing level annotations enrich the generation. A structured setup wizard, AI-assisted attribute suggestion, and live data analysis lower the technical barriers for researchers new to conjoint methodology. A full export bundle including all stimuli, their generating prompts, and response data facilitates transparency and reproducibility. The platform is demonstrated through a proof-of-concept study on care robot preferences for ambient assisted living (AAL, N=55) using AI-generated visual stimuli. The paper discusses the role of AI assistance in conjoint design, arguing that theoretical grounding must remain the researcher's responsibility, and outlining how genAI-generated stimuli can broaden the methodological repertoire for HCI and related fields.

Figures

Figures reproduced from arXiv: 2606.12972 by Philipp Brauner.

Figure 1
Figure 1. Figure 1: Screenshots of the GenerativeConjoint survey configuration wizard. [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: AI-generated stimuli (with ‘gpt-image-2’) from the robots for ambient assisted living study. Each image [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Attribute importance (top) and part-worth utilities (bottom) for the robots for ambient assisted living conjoint [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

30 extracted references · 24 canonical work pages

  1. [1]

    Who cares about crackdowns? Exploring the role of trust in individual philanthropy,

    S. Chaudhry, M. Dotson, and A. Heiss, “Who cares about crackdowns? Exploring the role of trust in individual philanthropy,”Global Policy, vol. 12, no. S5, pp. 45–58, 2021, doi: 10.1111/1758-5899.12984

    work page doi:10.1111/1758-5899.12984 2021

  2. [2]

    Using conjoint analysis to elicit preferences for health care,

    M. Ryan and S. Farrar, “Using conjoint analysis to elicit preferences for health care,” BMJ, vol. 320, no. 7248, pp. 1530–1533, 2000, doi: 10.1136/bmj.320.7248.1530

    work page doi:10.1136/bmj.320.7248.1530 2000

  3. [3]

    Mapping public perception of artificial intelli- gence: Expectations, risk–benefit tradeoffs, and value as determinants for societal acceptance,

    P . Brauner, F. Glawe, G. L. Liehner, L. V ervier, and M. Ziefle, “Mapping public perception of artificial intelli- gence: Expectations, risk–benefit tradeoffs, and value as determinants for societal acceptance,” Technological F orecasting and Social Change, vol. 220, p. 124304, 2025, doi: 10.1016/j.techfore.2025.124304

    work page doi:10.1016/j.techfore.2025.124304 2025

  4. [4]

    Assessing public acceptance of the life cycle of CO 2-based fuels: Does information make the difference?

    J. Offermann-van Heek, K. Arning, A. Sternberg, A. Bardow, and M. Ziefle, “Assessing public acceptance of the life cycle of CO 2-based fuels: Does information make the difference?” Energy Policy, vol. 143, p. 111586, 2020, doi: 10.1016/j.enpol.2020.111586

    work page doi:10.1016/j.enpol.2020.111586 2020

  5. [5]

    Conjoint analysis in consumer research: Issues and outlook,

    P . E. Green and V . Srinivasan, “Conjoint analysis in consumer research: Issues and outlook,” Journal of Con- sumer Research, vol. 5, no. 2, pp. 103–123, 1978, doi: 10.1086/208721

    work page doi:10.1086/208721 1978

  6. [6]

    B. K. Orme, Getting started with conjoint analysis: Strategies for product design and pricing research , 2nd ed. Research Publishers, 2010

    2010

  7. [7]

    Lighthouse studio

    Sawtooth Software, “Lighthouse studio.” Sawtooth Software, Inc., Provo, UT, 2024. Available: https://www. sawtoothsoftware.com/lighthouse-studio

    2024

  8. [8]

    Generating optimal designs for discrete choice ex- periments in R: The idefix package,

    F. Traets, D. G. Sanchez Arteta, and M. V andebroek, “Generating optimal designs for discrete choice ex- periments in R: The idefix package,” Journal of Statistical Software , vol. 96, no. 3, pp. 1–41, 2020, doi: 10.18637/jss.v096.i03

    work page doi:10.18637/jss.v096.i03 2020

  9. [9]

    Basic functions for supporting an implementation of choice experiments in R,

    H. Aizaki, “Basic functions for supporting an implementation of choice experiments in R,” Journal of Statisti- cal Software, vol. 50, no. 2, pp. 1–24, 2012, doi: 10.18637/jss.v050.c02

    work page doi:10.18637/jss.v050.c02 2012

  10. [10]

    On the design and analysis of simulated choice or allocation experiments in travel choice modelling,

    J. J. Louviere and D. A. Hensher, “On the design and analysis of simulated choice or allocation experiments in travel choice modelling,” Transportation Research Record, vol. 890, pp. 11–17, 1982

    1982

  11. [11]

    K. E. Train, Discrete choice methods with simulation , 2nd ed. Cambridge University Press, 2009. doi: 10.1017/CBO9780511805271

    work page doi:10.1017/cbo9780511805271 2009

  12. [12]

    W. F. Kuhfeld, Marketing research methods in SAS . SAS Institute, 2010

    2010

  13. [13]

    Individualized hybrid models for conjoint analysis,

    P . E. Green and A. M. Krieger, “Individualized hybrid models for conjoint analysis,” Management Science, vol. 42, no. 6, pp. 850–867, 1996, doi: 10.1287/mnsc.42.6.850

    work page doi:10.1287/mnsc.42.6.850 1996

  14. [14]

    M. Ryan, K. Gerard, and M. Amaya-Amaya, Eds., Using discrete choice experiments to value health and health care. Springer, 2008. doi: 10.1007/978-1-4020-5753-3

    work page doi:10.1007/978-1-4020-5753-3 2008

  15. [15]

    Conjoint analysis applications in health — a checklist: A report of the ISPOR Good Research Practices for Conjoint Analysis Task Force,

    J. F. P . Bridges et al. , “Conjoint analysis applications in health — a checklist: A report of the ISPOR Good Research Practices for Conjoint Analysis Task Force,” V alue in Health, vol. 14, no. 4, pp. 403–413, 2011, doi: 10.1016/j.jval.2010.11.013

    work page doi:10.1016/j.jval.2010.11.013 2011

  16. [16]

    Causal inference in conjoint analysis: Understanding multi- dimensional choices via stated preference experiments,

    J. Hainmueller, D. J. Hopkins, and T. Y amamoto, “Causal inference in conjoint analysis: Understanding multi- dimensional choices via stated preference experiments,” Political Analysis, vol. 22, no. 1, pp. 1–30, 2014, doi: 10.1093/pan/mpt024

    work page doi:10.1093/pan/mpt024 2014

  17. [17]

    The hidden American immigration consensus: A conjoint analysis of attitudes toward immigrants,

    J. Hainmueller and D. J. Hopkins, “The hidden American immigration consensus: A conjoint analysis of attitudes toward immigrants,” American Journal of Political Science , vol. 59, no. 3, pp. 529–548, 2015, doi: 10.1111/ajps.12138

    work page doi:10.1111/ajps.12138 2015

  18. [18]

    Users’ preferences for smart home automation— Investigating aspects of privacy and trust,

    E.-M. Schomakers, H. Biermann, and M. Ziefle, “Users’ preferences for smart home automation— Investigating aspects of privacy and trust,” Telematics and Informatics , vol. 64, p. 101689, 2021, doi: 10.1016/j.tele.2021.101689

    work page doi:10.1016/j.tele.2021.101689 2021

  19. [19]

    Which artificial intelligences do people care about most? A conjoint experiment on moral consideration,

    A. Ladak, J. Harris, and J. R. Anthis, “Which artificial intelligences do people care about most? A conjoint experiment on moral consideration,” in Proceedings of the CHI conference on human factors in computing systems, in CHI ’24. ACM, 2024, pp. 1–11. doi: 10.1145/3613904.3642403. 11 A P REPRINT - 2026-06-11

    work page doi:10.1145/3613904.3642403 2024

  20. [20]

    Appearance is not everything—Preferred feature combinations for care robots,

    K. Klüber and L. Onnasch, “Appearance is not everything—Preferred feature combinations for care robots,” Computers in Human Behavior , vol. 128, p. 107128, 2022, doi: 10.1016/j.chb.2021.107128

    work page doi:10.1016/j.chb.2021.107128 2022

  21. [21]

    Simkute, L

    M. Keil, L. V ervier, P . Brauner, and M. Ziefle, “Will you be watching me? A conjoint-based study on employee attitudes toward personal data usage in smart factories,”International Journal of Human–Computer Interaction, vol. 41, no. 16, pp. 10024–10044, 2024, doi: 10.1080/10447318.2024.2430494

    work page doi:10.1080/10447318.2024.2430494 2024

  22. [22]

    Simkute, L

    J. Y eon, Y . Jung, Y . Baek, D. Lee, J. Shin, and W. Y . Chung, “User preferences on a generative AI user interface through a choice experiment,” International Journal of Human–Computer Interaction , vol. 41, no. 12, pp. 7626–7637, 2024, doi: 10.1080/10447318.2024.2400379

    work page doi:10.1080/10447318.2024.2400379 2024

  23. [23]

    V erbal versus realistic pictorial representa- tions in conjoint analysis with design attributes,

    M. Vriens, G. H. Loosschilder, E. Rosbergen, and D. R. Wittink, “V erbal versus realistic pictorial representa- tions in conjoint analysis with design attributes,” Journal of Product Innovation Management , vol. 15, no. 5, pp. 455–467, 1998, doi: 10.1111/1540-5885.1550455

    work page doi:10.1111/1540-5885.1550455 1998

  24. [24]

    Modelling attribute non-attendance in choice experiments for rural landscape valuation,

    R. Scarpa, T. J. Gilbride, D. Campbell, and D. A. Hensher, “Modelling attribute non-attendance in choice experiments for rural landscape valuation,” European Review of Agricultural Economics , vol. 36, no. 2, pp. 151–174, 2009, doi: 10.1093/erae/jbp012

    work page doi:10.1093/erae/jbp012 2009

  25. [25]

    The effect of product representation in visual conjoint analysis,

    B. Sylcott, S. Orsborn, and J. Cagan, “The effect of product representation in visual conjoint analysis,” Journal of Mechanical Design, vol. 138, no. 10, 2016, doi: 10.1115/1.4034085

    work page doi:10.1115/1.4034085 2016

  26. [26]

    Auspurg and T

    K. Auspurg and T. Hinz, Factorial survey experiments , vol. 175. in Quantitative applications in the social sciences, vol. 175. SAGE, 2015. doi: 10.4135/9781483398075

    work page doi:10.4135/9781483398075 2015

  27. [27]

    A general method for constructing efficient choice designs,

    K. Zwerina, J. Huber, and W. F. Kuhfeld, “A general method for constructing efficient choice designs,” SAS Institute, 1996

    1996

  28. [28]

    D. K. V . Duyne, J. Landay, and J. I. Hong, The design of sites: Patterns, principles, and processes for crafting a customer-centered web experience. USA: Addison-Wesley Longman Publishing Co., Inc., 2002

    2002

  29. [29]

    Scoping review on the use of socially assistive robot technology in elderly care,

    J. Abdi, A. Al-Hindawi, T. Ng, and M. P . Vizcaychipi, “Scoping review on the use of socially assistive robot technology in elderly care,” BMJ Open, vol. 8, no. 2, p. e018815, 2018, doi: 10.1136/bmjopen-2017-018815

    work page doi:10.1136/bmjopen-2017-018815 2018

  30. [30]

    Design recom- mendations for socially assistive robots for health and social care based on a large scale analysis of stakeholder positions,

    H. L. Bradwell, G. E. Aguiar Noury, K. J. Edwards, R. Winnington, S. Thill, and R. B. Jones, “Design recom- mendations for socially assistive robots for health and social care based on a large scale analysis of stakeholder positions,” Health Policy and Technology, vol. 10, no. 3, p. 100544, 2021, doi: 10.1016/j.hlpt.2021.100544. 12

    work page doi:10.1016/j.hlpt.2021.100544 2021