pith. sign in

arxiv: 2604.26083 · v1 · submitted 2026-04-28 · 💻 cs.HC

Designing Rewards for Rewarding Designs: Demonstrating the Impact of Rewards on the Creative Design Process

Surabhi S Nath , Vindula Jayawardana , Monica Van , Matt Klenk , Shabnam Hakimi This is my paper

Pith reviewed 2026-05-07 15:26 UTC · model grok-4.3

classification 💻 cs.HC
keywords creative design processrewardsdesign space explorationgoal-aligned feedbackMarkov Decision Processiterative designhuman-computer interactionparametric modeling
0
0 comments X

The pith

In an iterative 3D chair design task, goal-aligned rewards prompt participants to explore the design space more thoroughly and favor those rewards while keeping output diversity intact.

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

The paper models a creative chair design activity as a sequence of decisions under constraints and tests how feedback in the form of rewards shapes what people do next. Participants built designs step by step and received either rewards that tracked progress toward the stated goal or rewards unrelated to it. Any form of reward increased the range of designs people tried, yet participants still worked hardest to collect the goal-aligned rewards. The concrete goal itself changed how useful people judged the rewards to be. These patterns suggest that reward choice can steer decision-making inside constrained creative work.

Core claim

By representing the parametric chair design activity as a Markov Decision Process, the study shows that participants who receive rewards at each step explore a larger portion of the possible design space than those who receive none, actively maximize the goal-aligned rewards over the goal-agnostic ones, and still produce sets of designs with comparable variety. The abstractness of the given goal further modulates how useful participants report the rewards to be.

What carries the argument

The Markov Decision Process representation of the 3D parametric chair task, in which each state is a chair configuration, each action changes one or more parameters, and a reward signal is delivered after every change.

If this is right

  • Participants explore a wider set of design options once rewards are introduced at each step.
  • People consistently work to collect goal-aligned rewards rather than goal-agnostic ones.
  • The variety of final designs remains similar whether rewards are present or absent.
  • The abstractness of the design goal changes how helpful participants find the reward signals.

Where Pith is reading between the lines

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

  • Reward signals could be added to interactive design software to encourage broader search without forcing convergence on a single solution.
  • The same reward logic might be tested in other sequential creative activities such as story writing or recipe development under fixed constraints.
  • Design tools could adapt the reward type dynamically once the goal's concreteness is known.

Load-bearing premise

The simplified 3D parametric chair task and the chosen reward signals capture the essential features of real-world creative design under constraints, and participants' self-reports accurately reflect how the rewards changed their decisions.

What would settle it

An experiment with the same reward structure but a different constrained creative task, such as iterative logo design or furniture layout, in which participants show no increase in design-space coverage and no preference for the goal-aligned reward.

Figures

Figures reproduced from arXiv: 2604.26083 by Matt Klenk, Monica Van, Shabnam Hakimi, Surabhi S Nath, Vindula Jayawardana.

Figure 1
Figure 1. Figure 1: Design environment with all features (within feature categories), sliders, and view at source ↗
Figure 2
Figure 2. Figure 2: A sample action sequence, corresponding states, and rewards in the design view at source ↗
Figure 3
Figure 3. Figure 3: Experimental procedure comprising three phases: practice, baseline (with view at source ↗
Figure 4
Figure 4. Figure 4: Example designs by goal designed by study participants in the baseline view at source ↗
Figure 5
Figure 5. Figure 5: Learned reward landscape with an example high-scoring design per goal. The view at source ↗
Figure 6
Figure 6. Figure 6: Effect of reward feedback on action space. (A) Distribution of number of view at source ↗
Figure 7
Figure 7. Figure 7: Reward maximisation by goal condition. (A) Distribution of rewards for view at source ↗
Figure 8
Figure 8. Figure 8: Participant ratings by reward-type (1-5) for (A) how much they referred view at source ↗
Figure 9
Figure 9. Figure 9: (A) Distributions of design diversity in practice, baseline, and reward phases. view at source ↗
Figure 10
Figure 10. Figure 10: Example high- and low-scoring designs for the goal-aligned reward. view at source ↗
read the original abstract

The creative design process involves transforming abstract goals into concrete outcomes through a series of decisions made under constraints. While such processes are commonly shaped by feedback like rewards, their impact on design decision making remains unclear. To better understand the role of rewards in the design process, we modeled a 3D parametric, goal-based chair design task as a Markov Decision Process. We tracked participants' decisions as they iteratively developed designs for an abstract design goal, and presented either a goal-aligned or goal-agnostic reward at every step. We tested the effect of these rewards on task behaviour and self-reported experience. With rewards, participants more thoroughly explored the design space, and maximised goal-aligned over goal-agnostic rewards while preserving diversity across designs. The nature of the goal also mattered, influencing participants' perception of the reward's usefulness. Building on these insights, we propose guidelines for designing effective feedback for design decision making.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

0 major / 2 minor

Summary. This paper examines the influence of rewards on the creative design process using a 3D parametric chair design task modeled as a Markov Decision Process (MDP). Participants iteratively refined designs for abstract goals while receiving either goal-aligned or goal-agnostic rewards at each decision step. Key findings indicate that the presence of rewards encourages more thorough exploration of the design space, participants tend to maximize goal-aligned rewards over goal-agnostic ones without sacrificing design diversity, and the specific nature of the design goal affects how useful participants perceive the rewards to be. The authors conclude by proposing guidelines for designing effective feedback mechanisms in design decision-making contexts.

Significance. Should the empirical results prove robust, this work contributes to the field of human-computer interaction by providing concrete evidence on how different reward structures shape exploration and decision-making in constrained creative tasks. The MDP modeling and decision tracking offer a replicable framework for studying design behaviors. The insights into goal-aligned versus agnostic rewards and the moderating effect of goal nature have direct implications for developing more effective interactive design tools and feedback systems.

minor comments (2)
  1. The abstract would benefit from including a brief mention of the sample size, statistical tests, and effect sizes to provide immediate context for the strength of the behavioral claims.
  2. The proposed guidelines for feedback design could be strengthened by linking them more explicitly to specific results from the user study, such as particular behaviors observed under different reward conditions.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for their supportive review, positive assessment of the work's significance, and recommendation for minor revision. We are pleased that the contributions to HCI, the replicable MDP framework, and implications for design tools were recognized.

Circularity Check

0 steps flagged

No significant circularity: empirical study with independent observations

full rationale

The manuscript describes an empirical user study in which participants performed an iterative 3D parametric chair design task modeled as an MDP; rewards (goal-aligned or goal-agnostic) were presented at each step and behavioral metrics plus self-reports were collected. No equations, fitted parameters, or predictions are claimed; the central results (greater exploration, preference for goal-aligned rewards, goal-nature effects) are direct summaries of observed data rather than reductions of any input by construction. No self-citation chains or uniqueness theorems appear in the load-bearing sections. The analysis is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Empirical behavioral study; no free parameters, mathematical axioms, or invented entities are introduced. All claims rest on experimental observations rather than derivations.

pith-pipeline@v0.9.0 · 5472 in / 1058 out tokens · 37090 ms · 2026-05-07T15:26:24.325382+00:00 · methodology

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

47 extracted references · 47 canonical work pages

  1. [1]

    B. R. Anderson, J. H. Shah, and M. Kreminski. Homogenization effects of large language models. In Creativity and Cognition , pages 413–425. ACM, 2024

    work page 2024

  2. [2]

    R. Bellman. A markovian decision process. Journal of Mathematics and Mechanics, 6(5):679–684, 1957

    work page 1957

  3. [3]

    D. E. Berlyne. Conflict, arousal, and curiosity . 1960. 19

    work page 1960

  4. [4]

    Bernal, J

    M. Bernal, J. R. Haymaker, and C. Eastman. Role of computational sup- port for designers. Design Studies , 41:163–182, 2015

    work page 2015

  5. [5]

    Busch, N

    S. Busch, N. S. Jensen, and M. Barros. Decoding design briefs. In Nordic Design Research Conference, 2023

    work page 2023

  6. [6]

    Cherry and C

    E. Cherry and C. Latulipe. Creativity support index. ACM TOCHI, 21:1– 25, 2014

    work page 2014

  7. [7]

    R. G. Cooper. The stage-gate idea to launch system, 2010

    work page 2010

  8. [8]

    Cristie and S

    V. Cristie and S. C. Joyce. Versioning for parametric design exploration. Automation in Construction , 129:103802, 2021

    work page 2021

  9. [9]

    N. Cross. Engineering design methods . 2021

    work page 2021

  10. [10]

    Dorst and N

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

    work page 2001

  11. [11]

    Fischer et al

    G. Fischer et al. Embedding computer-based critics in the contexts of design. In CHI ’93 . ACM, 1993

    work page 1993

  12. [12]

    J. S. Gero and U. Kannengiesser. Situated function-behaviour-structure framework. Design Studies , 25:373–391, 2004

    work page 2004

  13. [13]

    P. M. Gollwitzer and G. B. Moskowitz. Goal effects on action and cognition . 1996

    work page 1996

  14. [14]

    M. D. Hoffman and A. Gelman. The no-u-turn sampler. JMLR, 15:1593– 1623, 2014

    work page 2014

  15. [15]

    Ivcevic and M

    Z. Ivcevic and M. Grandinetti. Artificial intelligence as a tool for creativity. Journal of Creativity , 34:100079, 2024

    work page 2024

  16. [16]

    A. N. Kluger and A. DeNisi. Effects of feedback interventions on perfor- mance. Psychological Bulletin , 119:254–284, 1996

    work page 1996

  17. [17]

    Lahikainen, N

    J. Lahikainen, N. M. Ady, and C. Guckelsberger. Creativity and mdps, 2024

    work page 2024

  18. [18]

    Lebuda and M

    I. Lebuda and M. Benedek. Creative metacognition framework. Physics of Life Reviews , 46:161–181, 2023

    work page 2023

  19. [19]

    Lee et al

    C. Lee et al. Guicomp: A gui design assistant with real-time feedback. In CHI 2020 . ACM, 2020

    work page 2020

  20. [20]

    J. H. Lee and M. J. Ostwald. Creative decision-making in parametric design. Buildings, 10:242, 2020

    work page 2020

  21. [21]

    M. D. Lee and E. J. Wagenmakers. Bayesian cognitive modeling. Cambridge University Press, 2014

    work page 2014

  22. [22]

    S. W. Lee et al. Impact of sketch-guided vs prompt-guided 3d generative ais. In CHI Conference. ACM, 2024. 20

    work page 2024

  23. [23]

    Leue and A

    A. Leue and A. Beauducel. Reinforcement sensitivity theory meta-analysis. Personality and Social Psychology Review , 12, 2008

    work page 2008

  24. [24]

    D. C.-E. Lin et al. Inkspire: Supporting design exploration with generative ai. In CHI Conference. ACM, 2025

    work page 2025

  25. [25]

    Y.-C. Liu, A. Chakrabarti, and T. Bligh. Ideal approach for concept gen- eration. Design Studies , 24:341–355, 2003

    work page 2003

  26. [26]

    Nandy et al

    A. Nandy et al. Semantic properties of word prompts shape design out- comes. In Design Computing and Cognition . Springer, 2025

    work page 2025

  27. [27]

    Nandy and K

    A. Nandy and K. Goucher-Lambert. How does machine advice influence de- sign choice? In Design Computing and Cognition , pages 801–818. Springer, 2023

    work page 2023

  28. [28]

    S. S. Nath, P. Dayan, and C. Stevenson. Characterising the creative process,

    work page

  29. [29]

    S. S. Nath et al. Relating objective complexity and beauty. Psychology of Aesthetics, Creativity, and the Arts , 2024

    work page 2024

  30. [30]

    A. Ng, D. Harada, and S. Russell. Policy invariance under reward transfor- mations. In ICML, pages 278–287, 1999

    work page 1999

  31. [31]

    Niv et al

    Y. Niv et al. Tonic dopamine. Psychopharmacology, 191:507–520, 2007

    work page 2007

  32. [32]

    A. Pan, K. Bhatia, and J. Steinhardt. Effects of reward misspecification,

    work page

  33. [33]

    J. K. Pugh, L. B. Soros, and K. O. Stanley. Quality diversity. Frontiers in Robotics and AI , 3, 2016

    work page 2016

  34. [34]

    R. M. Ryan. Control and information in the intrapersonal sphere. Journal of Personality and Social Psychology , 43:450–461, 1982

    work page 1982

  35. [35]

    R. M. Ryan and E. L. Deci. Self-determination theory. American Psychol- ogist, 55:68–78, 2000

    work page 2000

  36. [36]

    D. A. Schön. Designing as reflective conversation. Knowledge-Based Sys- tems, 5:3–14, 1992

    work page 1992

  37. [37]

    D. A. Schon and V. DeSanctis. The reflective practitioner: How profes- sionals think in action. Journal of Continuing Higher Education , 34:29–30, 1986

    work page 1986

  38. [38]

    Shireen et al

    N. Shireen et al. Design space exploration in parametric systems. In Cre- ativity and Cognition . ACM, 2011

    work page 2011

  39. [39]

    H. A. Simon. The structure of ill structured problems. Artificial Intelligence, 4:181–201, 1973

    work page 1973

  40. [40]

    Son et al

    K. Son et al. Creativesearch: Proactive design exploration system. Au- tomation in Construction , 142:104502, 2022. 21

    work page 2022

  41. [41]

    Son et al

    K. Son et al. Genquery: Supporting expressive visual search. In CHI Conference. ACM, 2024

    work page 2024

  42. [42]

    K. Son, K. Kim, and K. H. Hyun. Bigexplore: Bayesian information gain framework. In CHI Conference. ACM, 2022

    work page 2022

  43. [43]

    R. S. Sutton and A. G. Barto. Reinforcement learning: An introduction . MIT Press, 1998

    work page 1998

  44. [44]

    Swearngin et al

    A. Swearngin et al. Scout: Rapid exploration of interface layout alterna- tives. In CHI 2020 . ACM, 2020

    work page 2020

  45. [45]

    S. G. Valeri et al. Implementation of the phase review process in new product development: A successful experience, 2003

    work page 2003

  46. [46]

    Wadinambiarachchi et al

    S. Wadinambiarachchi et al. Effects of generative ai on design fixation. In CHI Conference. ACM, 2024

    work page 2024

  47. [47]

    M. B. Waldron and K. J. Waldron. Influence of designer expertise. In Mechanical Design: Theory and Methodology , pages 5–20. Springer, 1996. 22

    work page 1996