pith. sign in

arxiv: 2604.27812 · v1 · submitted 2026-04-30 · 💻 cs.HC

"It depends on where AI is used": Players' attitude patterns and evaluative logics toward different AI applications in digital games

Pith reviewed 2026-05-07 05:57 UTC · model grok-4.3

classification 💻 cs.HC
keywords AI in gamesplayer attitudesevaluative logicsthematic analysisdigital gamesAI acceptanceimmersionautonomy
0
0 comments X

The pith

Players accept AI in digital games only when it enhances immersion or efficiency without threatening creativity or autonomy

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

The paper examines how players evaluate eight specific AI uses in games, including intelligent NPCs, emergent narratives, dynamic balancing, recommendation systems, review processes, art generation, co-creation, and gameplay evolution. Based on nearly 1,900 open-ended responses from 310 players, it finds systematic patterns: AI gains support for boosting immersion, personalization, novelty, efficiency, and convenience, but faces resistance when it risks creativity, emotional authenticity, autonomy, fairness, stability, authorship, or accountability. The responses cluster into six evaluative logics that organize player reasoning. A sympathetic reader cares because broad pro-AI or anti-AI positions overlook these context-specific reactions that shape actual adoption. If the patterns hold, developers gain a practical map for choosing which AI features to implement and how to frame them to players.

Core claim

Players welcomed AI when it enhanced immersion, personalization, novelty, efficiency, or convenience, but resisted it when it threatened creativity, emotional authenticity, autonomy, fairness, system stability, authorship, or accountability. Responses organized around six evaluative logics: experiential enrichment, instrumental efficiency, system reliability, agency and control, authorship and compliance, and human oversight.

What carries the argument

Thematic analysis of open-ended responses to eight distinct AI application contexts, which surfaces six evaluative logics that players apply when judging acceptance or rejection

Load-bearing premise

That thematic analysis of self-reported open-ended responses from a sample of 310 players accurately captures generalizable evaluative logics without significant researcher bias or limitations from convenience sampling

What would settle it

A larger, more diverse survey or in-game experiment in which player attitudes toward the same eight AI contexts do not cluster into the six logics and instead show uniform acceptance or rejection regardless of application

Figures

Figures reproduced from arXiv: 2604.27812 by Fei Qin, Jiangxu Lin, Ting-Chen Hsu, Wenran Chen, Zheyuan Zhang.

Figure 2
Figure 2. Figure 2: integrates these findings by linking the eight application contexts to broader sites of AI intervention and to the six evaluative logics. The model shows that players' attitudes depend less on AI use in games in general than on where AI is introduced, what role it performs, and which values it affects. Player acceptance or resistance can therefore be understood as a context￾sensitive evaluation of AI's exp… view at source ↗
read the original abstract

As AI becomes increasingly embedded in digital games, players' attitudes de-pend not only on whether AI is used, but also on where and how it intervenes in gameplay. This study examines players' evaluative patterns toward eight AI application contexts, including intelligent NPCs, emergent narrative, dynamic balancing, recommendation systems, review and governance, art asset generation, co-creation gameplay, and gameplay evolution. Based on 1,856 valid open-ended responses from 310 questionnaires, we conducted thematic analysis to identify reasons for acceptance, rejection, and conditional acceptance. Results show that players welcomed AI when it enhanced immersion, personalization, novelty, efficiency, or convenience, but resisted it when it threatened creativity, emotional authenticity, autonomy, fairness, system stability, authorship, or accountability. We further identify six evaluative logics: experiential enrichment, instrumental efficiency, system reliability, agency and control, authorship and compliance, and human oversight. These preliminary findings highlight the context-sensitive nature of AI acceptance in digital games.

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

2 major / 3 minor

Summary. This paper investigates players' attitudes toward AI in digital games across eight specific application contexts using a survey of 310 participants that collected 1,856 open-ended responses. Through thematic analysis, it categorizes reasons for acceptance (e.g., enhanced immersion, personalization, novelty, efficiency) and rejection (e.g., threats to creativity, emotional authenticity, autonomy, fairness, authorship), and proposes six evaluative logics: experiential enrichment, instrumental efficiency, system reliability, agency and control, authorship and compliance, and human oversight. The key finding is that acceptance depends on the specific use of AI rather than AI in general.

Significance. If the results hold, the paper makes a meaningful contribution to HCI and game studies by moving beyond binary views of AI acceptance to a context-sensitive framework of evaluative logics. The empirical basis with a reasonable sample size and focus on diverse AI applications (from NPCs to art generation) provides actionable insights for game designers and AI developers. It highlights important player concerns around agency and authenticity that could influence ethical AI deployment in entertainment media.

major comments (2)
  1. [Methods] The thematic analysis is not described with sufficient rigor. No details are provided on the coding process, including codebook creation, involvement of multiple researchers, inter-rater reliability statistics, or saturation assessment. This is a load-bearing issue for the central claim, as the six evaluative logics are directly derived from this analysis of the 1,856 responses, and without these elements, it is difficult to rule out researcher bias in theme identification.
  2. [Methods] The participant recruitment and sample characteristics are insufficiently detailed. While convenience sampling is implied, there is no discussion of potential biases (such as overrepresentation of dedicated gamers from specific online communities) or how this might influence the emergence of logics like 'agency and control' and 'authorship and compliance'. This affects the generalizability of the findings.
minor comments (3)
  1. [Abstract] Typographical error: 'de-pend' should read 'depend'.
  2. [Results] To strengthen the presentation, include representative verbatim quotes from participants for each of the six evaluative logics to demonstrate how the themes were identified from the data.
  3. Consider adding a table summarizing the acceptance/rejection reasons across the eight AI contexts for better readability.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive and positive review, which highlights the paper's potential contribution while identifying areas for methodological improvement. We agree with the comments and will revise the manuscript to enhance transparency and rigor in the Methods section.

read point-by-point responses
  1. Referee: [Methods] The thematic analysis is not described with sufficient rigor. No details are provided on the coding process, including codebook creation, involvement of multiple researchers, inter-rater reliability statistics, or saturation assessment. This is a load-bearing issue for the central claim, as the six evaluative logics are directly derived from this analysis of the 1,856 responses, and without these elements, it is difficult to rule out researcher bias in theme identification.

    Authors: We appreciate the referee's emphasis on methodological transparency, as the thematic analysis underpins the six evaluative logics. The original manuscript provided a high-level overview but lacked the requested specifics. In the revision, we will expand the Methods section to describe: iterative codebook development via open coding on an initial subset of responses; involvement of two researchers who independently coded all 1,856 responses with weekly consensus discussions to resolve discrepancies; inter-rater reliability assessed via Cohen's kappa on a 20% random sample (targeting >0.75); and thematic saturation evaluation, which was reached after coding approximately 65% of responses with no new themes emerging. These details will strengthen confidence in the derivation of the logics and address potential bias concerns. revision: yes

  2. Referee: [Methods] The participant recruitment and sample characteristics are insufficiently detailed. While convenience sampling is implied, there is no discussion of potential biases (such as overrepresentation of dedicated gamers from specific online communities) or how this might influence the emergence of logics like 'agency and control' and 'authorship and compliance'. This affects the generalizability of the findings.

    Authors: We agree that fuller disclosure of recruitment and sample limitations is needed to contextualize the findings. We will revise the Methods section to detail the convenience sampling approach (primarily via Reddit gaming subreddits, Discord servers, and university networks) and include comprehensive demographics (e.g., age distribution, gender, weekly gaming hours, and self-reported expertise). A new Limitations subsection will explicitly discuss self-selection bias toward dedicated gamers, noting that this may amplify logics such as agency and control or authorship and compliance. We will also suggest that future work with more representative samples could test broader applicability, thereby improving the paper's transparency on generalizability without overstating the current results. revision: yes

Circularity Check

0 steps flagged

No circularity: central claims derived from primary data collection and inductive thematic analysis

full rationale

The paper collects original questionnaire responses from 310 participants yielding 1,856 open-ended answers, then applies thematic analysis to surface acceptance/rejection reasons and six evaluative logics. No equations, fitted parameters, self-citations, or prior-work ansatzes are invoked to generate these results; the logics and patterns are presented as emerging directly from the new data. The derivation is therefore self-contained and does not reduce to its own inputs by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The paper does not introduce free parameters or invented entities. It relies on standard assumptions in qualitative HCI research about the validity of self-reported attitudes and the reliability of thematic analysis for identifying underlying logics.

axioms (1)
  • domain assumption Self-reported attitudes in open-ended survey responses can be thematically analyzed to reliably reveal distinct evaluative logics without substantial loss or distortion of meaning.
    This assumption is invoked when mapping the 1,856 responses to the six evaluative logics of experiential enrichment, instrumental efficiency, system reliability, agency and control, authorship and compliance, and human oversight.

pith-pipeline@v0.9.0 · 5486 in / 1582 out tokens · 70235 ms · 2026-05-07T05:57:06.852456+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

80 extracted references · 80 canonical work pages · 1 internal anchor

  1. [1]

    In: Proceedings of the 30th ACM Symp osium on Virtual Reality Software and Technology, pp

    Christiansen, F.R., Hollensberg, L.N., Jensen, N.B., Julsgaard, K., Jespersen, K.N., Nikolov, I.: Exploring presence in interactions with LLM -driven NPCs: a comparative study of speech recognition and dialogue options. In: Proceedings of the 30th ACM Symp osium on Virtual Reality Software and Technology, pp. 1–11 (2024)

  2. [2]

    In: Findings of the Association for Computational Lingu istics: ACL 2025, pp

    SongTang, S., Zhao, K., Wang, L., Li, Y., Liu, X., Zou, J., Wang, Q., Chu, X.: UnrealLLM: Towards highly controllable and interactable 3D scene generation by LLM-powered proce- dural content generation. In: Findings of the Association for Computational Lingu istics: ACL 2025, pp. 19417–19435 (2025)

  3. [3]

    I Like Your Story!

    Fu, J., Wang, X., Vi, K., Li, Z., Xu, C., Sun, Y.: “I Like Your Story!”: A co -creative story- crafting game with a persona -driven character based on generative AI. In: Proceedings of the Extended Abstracts of the CHI Conference on Human Factors in Computin g Systems, pp. 1–5 (2025)

  4. [4]

    In: Proceedings of the Extended Abstracts of the CHI Conference on Human Factors in Computing Systems, pp

    Jeong, J., Lee, T.Y.: Ligs: developing an LLM-infused game system for emergent narrative. In: Proceedings of the Extended Abstracts of the CHI Conference on Human Factors in Computing Systems, pp. 1–12 (2025)

  5. [5]

    Alharthi, S.A.: Generative AI in game design: Enhancing creativity or constraining innova- tion? Journal of Intelligence 13(6), 60 (2025)

  6. [6]

    In: Proceedings of the 2026 CHI Conference on Human Factors in Com- puting Systems, pp

    Bazzaz, M., Cooper, S.: Playing the imitation game: How perceived generated content shapes player experience. In: Proceedings of the 2026 CHI Conference on Human Factors in Com- puting Systems, pp. 1–21 (2026)

  7. [7]

    In: Proceedings of the 31st International Con- ference on Intelligent User Interfaces, pp

    Hochreiter, M., Kriglstein, S., Wallner, G.: Beyond pre -defined scripts: Player perceptions on generative non-player character dialogues. In: Proceedings of the 31st International Con- ference on Intelligent User Interfaces, pp. 2004–2018 (2026)

  8. [8]

    In: Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems, pp

    Milani, S., Juliani, A., Momennejad, I., Georgescu, R., Rzepecki, J., Shaw, A., Costello, G., Fang, F., Devlin, S., Hofmann, K.: Navigates like me: Understanding how people evaluate human-like AI in video games. In: Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems, pp. 1–18 (2023)

  9. [9]

    The Double-Edged Sword of Open-Ended Interaction: How LLM-Driven NPCs Affect Players' Cognitive Load and Gaming Experience

    Hsu, T.C., Chen, W., Lin, J., Qin, F., Zhang, Z.: The double -edged sword of open -ended interaction: How LLM-driven NPCs affect players' cognitive load and gaming experience. arXiv preprint arXiv:2604.10107 (2026). 10 Author Appendix A Questionnaire Design (1) What is your nickname? (Fill in the blank) ______ (2) What is your age? (Fill in the blank, an ...

  10. [10]

    Enhancing immersion/presence: Players believe AI-NPCs can effectively enhance the gaming experience, make characters feel more realistic, increase engagement and investment in the game, and make players more willing to participate

  11. [11]

    High personalization/richness: Players believe AI-NPCs can create personalized experiences, enrich game content, make experiences more diverse, and bring greater freshness

  12. [12]

    High logicality: Players believe AI-NPCs produce replies with strong logic, high quality, and clear organization

  13. [13]

    High freedom: Players believe AI -NPCs significantly increase freedom and openness in the game, allowing players to express ideas more freely

  14. [14]

    Reducing development costs: Players believe that developing AI-driven NPCs can effectively reduce workload for developers, improve development efficiency, reduce effort spent on creating NPC settings, and decrease the need for writing dialogue trees

  15. [15]

    Logical confusion: Players believe AI -NPC replies often show illogical connections, incon- sistencies, irrelevant answers, or nonsensical content

  16. [16]

    Lack of emotion: Players believe AI -NPCs lack emotional depth, feel robotic or artificial, become boring over time, and make it difficult to empathize

  17. [17]

    Out of character: Players believe AI -NPCs easily go out of character (OOC), show incon- sistent personality or reply style, or have fragmented continuity in responses

  18. [18]

    Low system stability: Players believe adding AI -NPCs may increase overall game system instability, such as causing bugs, progression blocks, unresponsive NPCs, deviation from main quests, or other system disruptions. Scenario 2: Attitudes toward AI for Dynamic/Emergent Narrative and Task Generation [8 themes] Themes: [Lack of creativity/emotion]; [Low sy...

  19. [19]

    Lack of creativity/emotion: Players feel AI -generated narratives lack human creative input, originality, or emotional depth; stories feel formulaic or players simply resist fully AI-generated plots

  20. [20]

    Low system stability: Players worry AI narrative generation may cause bugs, progression blocks, or task errors

  21. [21]

    Enhancing immersion: Players believe AI -generated narratives effectively deepen game ex- perience and engagement

  22. [22]

    12 Author

    High personalization/richness: Players believe AI can meet various narrative preferences, in- crease content richness and variety, and offer more branches or possibilities. 12 Author

  23. [23]

    High replayability: Players believe it improves gameplay value, increases replayability, and reduces repetition

  24. [24]

    Chaotic/poor -quality narrative: Emphasizes issues with internal logic, deviation from the main storyline or world-setting, and overall low narrative quality (distinct from system stability issues)

  25. [25]

    Strong novelty/freshness: Players expect greater freshness, higher anticipation, stronger ex- ploration desire, varied experiences per playthrough, lower repetition, more randomness and sur- prise, higher openness, and more diverse endings

  26. [26]

    Extending game lifespan: Players explicitly mention that AI -generated narratives make the game more enduring and extend its lifecycle. Scenario 3: Attitudes toward AI for Dynamic Game Difficulty/Balance Adjustment [9 themes] Themes: [Enhanced personalization/adaptability]; [Reducing balancing costs]; [Unstable balancing]; [Dis- rupting game rhythm]; [Enh...

  27. [27]

    Enhanced personalization/adaptability: AI can meet the needs of players at different skill levels and create personalized experiences

  28. [28]

    Reducing balancing costs: AI lowers the design effort and repeated manual adjustments re- quired for game balance, improving efficiency

  29. [29]

    Unstable balancing: AI adjustments may produce unreasonable or inconsistent balance, inac- curacies, or bugs

  30. [30]

    Disrupting game rhythm: AI interference breaks flow state, planned strategies, and overall game feel, turning engaging gameplay monotonous

  31. [31]

    Enhancing immersion: AI adjustments improve player engagement and willingness to partic- ipate

  32. [32]

    hand-holding,

    Impaired autonomy: Players want to choose their own difficulty, feel their self -efficacy is undermined by “hand-holding,” or prefer to attribute difficulty solely to their own skill

  33. [33]

    Reducing frustration: Helps lower -skill players avoid frustration and reduces dropout rates (new-player friendly)

  34. [34]

    Maintaining challenge: Prevents the game from becoming too easy for high -skill players, maintaining motivation through balanced challenge

  35. [35]

    Reduced sense of achievement: Dynamic easing makes victories feel unearned or diminishes the reward of overcoming difficulty. Scenario 4: Attitudes toward AI for Personalized Matching or Recommendation Systems [7 themes] Themes: [Prone to monotony]; [Inaccurate matching]; [Enhancing gaming experience]; [Improving gam- ing efficiency]; [Privacy concerns]; ...

  36. [36]

    Prone to monotony: Long-term recommendations based on past behavior create filter bubbles, aesthetic fatigue, and lack of novelty

  37. [37]

    Inaccurate matching: AI fails to accurately understand player preferences, abilities, or char- acteristics

  38. [38]

    Enhancing gaming experience: Improves immersion, engagement, interaction quality, bal- ance, fairness, etc

  39. [39]

    Improving gaming efficiency: Saves time and mental effort in selecting content or opponents

  40. [40]

    It depends on where AI is used 13

    Privacy concerns: Fear of personal data collection and potential leaks. It depends on where AI is used 13

  41. [41]

    Catering to diverse preferences: Effectively meets varied player needs and delivers unique personalized experiences

  42. [42]

    Impaired autonomy: Players resist being overly guided or having choices made for them. Scenario 5: Attitudes toward AI for Report Review or Community Governance [7 themes] Themes: [Lack of emotional feedback]; [Strong fairness/objectivity]; [Missed/misjudgment concerns]; [Bias/fairness concerns]; [High efficiency]; [Need for human-AI collaboration]; [Lack...

  43. [43]

    Lack of emotional feedback: AI responses feel robotic, insincere, or perfunctory

  44. [44]

    Strong fairness/objectivity: AI delivers logical, consistent, and fair rulings in most cases

  45. [45]

    Missed/misjudgment concerns: AI may be unstable, miss violations, or wrongly classify con- tent

  46. [46]

    Bias/fairness concerns: AI may systematically favor certain player groups

  47. [47]

    High efficiency: Much faster processing and timely feedback than human moderation

  48. [48]

    Need for human-AI collaboration: Players prefer a hybrid approach rather than full AI or full human control

  49. [49]

    Lack of semantic/emotional understanding: AI struggles with complex human behavior, sar- casm, slang, or nuanced social contexts. Scenario 6: Attitudes toward AI for Art Asset Creation [12 themes] Themes: [AI-assisted & human -led]; [Copyright risks]; [Weak innovation/creativity]; [Low aesthetics]; [High aesthetics]; [Poor controllability]; [Benefiting de...

  50. [50]

    AI-assisted & human -led: AI can provide references or inspiration, but core creative work should remain human-driven

  51. [51]

    Copyright risks: Concerns about infringement or unclear legal status of AI-generated assets

  52. [52]

    AI -like,

    Weak innovation/creativity: Outputs feel generic, “AI -like,” and lack the spark of human creativity

  53. [53]

    Low aesthetics: Reduces visual/auditory quality and causes aesthetic downgrade

  54. [54]

    High aesthetics: Delivers superior visual/auditory quality compared to traditional methods

  55. [55]

    Poor controllability: Frequent hallucinations, anatomical errors, style inconsistency, etc

  56. [56]

    Benefiting developers: Accelerates iteration, lowers costs, and improves development effi- ciency

  57. [57]

    Lack of creator empathy: Players cannot emotionally connect with or feel the human artistic intent behind the work

  58. [58]

    Low generation quality: Assets look bad, sound bad, or feel uninteresting

  59. [59]

    High generation quality: Assets look good, sound good, and feel high-quality

  60. [60]

    Lack of authenticity/emotion: Images feel unreal, music/voice lacks soul, etc

  61. [61]

    Strong perfunctory feeling: Use of AI assets feels disrespectful to players, as if the develop- ers cut corners. Scenario 7: Attitudes toward AI Supporting Player Co -Creation of Personalized Content [11 themes] Themes: [Copyright/compliance concerns]; [Lowering creation threshold]; [Meeting diverse demands]; [Improving creation efficiency]; [Enhancing im...

  62. [62]

    Copyright/compliance concerns: Risk of infringement, violation of platform rules, ethics, or law

  63. [63]

    Lowering creation threshold: Enables players with limited skills to realize their ideas

  64. [64]

    Meeting diverse demands: Allows realization of varied personal visions and unique experi- ences

  65. [65]

    Improving creation efficiency: Saves time realizing ideas

  66. [66]

    Enhancing immersion: Increases engagement and investment in the game

  67. [67]

    Enhancing novelty/richness: Continuously introduces new, unique elements

  68. [68]

    Quality concerns: Generated content may be low-quality or fail to meet expectations

  69. [69]

    High freedom/openness: Expands game boundaries and player creative freedom

  70. [70]

    Lack of authorship: Players feel the creations are not truly “theirs.”

  71. [71]

    Meeting sense of achievement: The creation process provides strong accomplishment

  72. [72]

    Meeting sharing desire: Encourages players to share their creations. Scenario 8: Attitudes toward AI for Emergent/Dynamic Gameplay Mechanisms [8 themes] Themes: [Meeting diverse demands]; [Content quality concerns]; [Disrupting balance/system stability]; [Improving development efficiency]; [Increasing cognitive load]; [Enhancing immersion/explo- ration de...

  73. [73]

    Meeting diverse demands: Realizes many player-desired mechanics and personalized experi- ences

  74. [74]

    Content quality concerns: Emergent elements may be low-quality or poorly implemented

  75. [75]

    Disrupting balance/system stability: New elements break designed balance, world tone, or cause bugs and chaos

  76. [76]

    Improving development efficiency: Reduces developer workload for new content, mods, or expansions

  77. [77]

    Increasing cognitive load: Constant new elements cause fatigue and make deep system mas- tery difficult

  78. [78]

    Enhancing immersion/exploration desire: Encourages deeper engagement and motivation to explore

  79. [79]

    Increasing novelty/richness: Provides ongoing fresh experiences and greater content depth

  80. [80]

    Extending game lifespan: Makes the game more enduring and worthy of long-term play