The integration of Generative AI into creative processes presents both opportunities and challenges for human creativity. While AI can augment creative capabilities, over-reliance can risk diminishing creative self-beliefs and human agency. This paper introduces a friction-by-design approach through the DA VINCI model, proposing purposeful micro-interventions that slow human-AI interactions at critical decision points. We present a frictionby-design approach to each mental state of the creative process, designed to prioritize human agency while leveraging AI capabilities that embodies context-adaptive friction to support meaningful human-AI co-creation.
Creativity, defined as the contemporaneous existence of potential originality and efectiveness which
may or may not turn into creative achievement [
Current research identifies multiple scenarios for cyber-creative processes [
CEUR
ceur-ws.org
However, Gen-AI tools usage may systematically diminish these creative self-beliefs [
To address these challenges, we introduce a ´´friction-by-design” approach to cyber-creative processes
within the DA VINCI model framework [
Our approach aims at fostering more conscious and meaningful human-AI interactions, ultimately enhancing creative engagement and agency while mitigating risks of over-reliance. By strategically decelerating the cyber-creative process at key junctures, we intend to enable fuller expression of individual creative potential.
In this article, we present practical implementation guidelines through the DA VINCI model for
cyber-creative processes, specifically focusing on a friction-by-design approach. Originally designed to
support human creativity, the DA VINCI model has been extended in 2023 as a framework for
Co-cre-AItion. This model employs a five-part acronym identifying key mental states in creative processes: DAV
(Drive, Attention & Volition) emphasizes energy investment, attention, and motivation in initiating
creativity; I (Information) encompasses both gathering of relevant knowledge and the introduction of
apparently irrelevant sources of inspiration; N (Novelty generation) focuses on idea creation with both
convergent and divergent modalities; C (Creativity estimation) involves evaluating generated ideas’
potential value, with an open mind towards serendipity; and I (Implementation) addresses creative
concept realization. The model’s dynamic structure allows iterative movement between diferent states
[
As indicated before, the DA VINCI model structures creative cognition into five interlinked mental states:
DAV—Drive, Attention & Volition; I—Information; N—Novelty generation; C—Creativity estimation;
I—Implementation. Each state may incorporate specific micro-interventions of the cyber-creative
frictions, stimulating deliberate design choices, fostering reflection and cognitive engagement, while
potentially reducing processing speed. These programmed hurdles aim to stimulate human cognitive
activation and mitigate over-reliance on automated systems [
In Table 1 the friction-by-design approach is presented in a unified way. These are the questions that form the micro-interventions aiming at preserving human agency.
The core of this mental state is the entry point into the creative process, characterized by the necessary investment of energy and time beyond basic survival levels. It is built on both cognitive (Attention) and motivational (Volition) elements. Attention focuses on a problem, while volition provides the crucial motivational drive.
Drive friction may prompt an individual to question why they are pursuing a specific path and whether it is truly meaningful or simply convenient. This friction can invite the consideration of multiple directions, not just the easiest, and reinforce a sense of agency in defining one’s creative trajectory.
DAV (Drive: Attention & Volition)
Drive friction I (Information)
Information friction N (Novelty Generation)
Generation friction C (Creativity estimation)
Estimation friction I (Implementation)
Implementation friction
What are multiple refined focused areas (RFAs), or directions, I could pursue — not just the easiest one? Am I choosing this path because it’s meaningful, or just because it’s convenient? Do I feel agency in defining my own creative direction? Have I verified the information, or am I accepting it just because it’s given (by AI or system)? Are my sources diverse, reliable, and relevant? What happens if I include irrelevant, absurd, or paradoxical information — does it spark new insights? Does this input challenge my assumptions, or just reinforce them? Can I generate ideas independently before seeing suggestions? How might an unexpected constraint reshape this idea? Can I form unusual associations between concepts, even if they seem illogical? Which idea surprised me most — and why? Why am I choosing one idea over another — can I justify it? Can I extract hidden potential from a weak idea by reframing it in another context? What would this idea look like from a completely diferent perspective? Am I critically assessing or just defaulting to the “most polished” suggestion? Have I iterated enough, or am I accepting the first viable version? At each stage, do I approve, refine, and adjust — or just let it pass? What feedback or reflection checkpoints can I add before finalizing? Does this outcome still align with my original motivation and values?
This state provides the essential elements for creative thinking, consisting of two components: Relevant Information from existing knowledge, and Inspiration, which includes information that might seem irrelevant or absurd but can spark non-linear thinking. Together, these components form the mental ´´Platform” for new ideas.
Information friction might serve as a mechanism to reduce misleading or ungrounded outcomes like AI hallucinations. It encourages a critical verification of information and a deliberate openness to unconventional viewpoints that challenge assumptions, thereby increasing the potential for originality.
This is the core of the creative process, where novel ideas are generated from the platform. It employs two thinking styles: convergent, which uses available elements synergistically to allow a new idea to emerge, and divergent, which requires generating multiple alternatives from a common root.
Generation friction may push the human to generate ideas independently before seeing suggestions. It encourages the use of unexpected constraints and the formation of unusual, illogical associations to reshape ideas and preserve the human’s role and sense of ownership in the creative process.
This state is where the value of generated ideas is estimated. For a truly original idea, its value cannot be judged statically but must be estimated by projecting it into a vision of the future. This is done through both convergent estimation, which assesses relevance to the initial problem, and divergent estimation, which explores value outside the original boundaries.
Estimation friction is intended to compel agents to justify their choices and critically assess ideas, rather than defaulting to the most risk-free suggestion. It encourages extracting hidden potential from a ´´weak” idea by reframing it from diferent perspectives.
This state is the final action-oriented phase where a creative episode becomes a visible, audible, or tangible result. It involves selecting and refining ideas that align with practical constraints and transforming them into early prototypes to be tested in the real world. It is the pathway that bridges creativity into innovation.
Implementation friction might encourage a creator to question whether they have refined their idea enough or are accepting the first viable version. It prompts the addition of reflection checkpoints before ifnalizing and ensures the outcome still aligns with the original motivation and values.
This friction-by-design approach framework aims to clarify friction introduction points and
significance, in order to align cyber-creative trajectories with human goals rather than speed and optimization
by default [
Future work will focus on implementing the friction approach into a customized GPT system, DA VINCI
2.0, building on the previous DA VINCI custom GPT [
In addition, mid-session recalibration is planned to occur after the Novelty stage through targeted queries about pace preference and exploration depth. This step acknowledges that optimal friction levels may evolve as creative work progresses and as users gain clarity about their goals. Sessions conclude with momentum-preserving prompts that encourage immediate action while leaving pathways for future refinement.
Once calibrated, DA VINCI 2.0 could dynamically monitor engagement patterns to sustain creative lfow, reducing resistance when users exhibit signs of overload. By combining structured calibration with adaptive modulation during interaction, DA VINCI 2.0 could serve as a future model for how generative systems balance eficiency with deliberate resistance, supporting creativity while maintaining human agency.
Furthermore, a promising avenue for future work lies in refining and expanding the current set of frictions. While we introduce one primary friction per mental state, it is possible that multiple, more ifne-grained frictions could exist within each stage of the creative process. For instance, the broad category of “assessment friction” might be further decomposed into frictions targeting self-evaluation, peer feedback, or long-term value reflection. Such reconfigurations may help tailor friction-by-design interventions to specific contexts, user profiles, or creative domains.
This paper describes an approach to the challenges of human-AI collaboration in creative processes through the introduction of systematic friction design. The DA VINCI model framework, combined with a context-adaptive friction approach, provides a structured methodology intended to maintain human agency while drawing on AI capabilities in cyber-creative settings.
The friction-by-design approach is positioned as a response to a gap in current human-AI interaction paradigms, where decision-making processes are intentionally slowed at certain points. Instead of emphasizing speed and eficiency, the proposed frictions are oriented toward reflective engagement, cognitive activation, and the support of creative self-beliefs. The DA VINCI model structure is outlined as a framework through which these interventions may be placed across the creative process.
The implications of this work are framed as extending beyond individual creative tools to broader considerations of human-AI collaboration design. As AI systems become more sophisticated and present across creative domains, concerns around diminished human agency and creative self-eficacy appear more relevant. The approach outlined here positions purposeful ineficiencies and reflective pauses not as limitations, but as features that may help preserve human creativity in the context of AI integration.
This work was supported by the National Recovery and Resilience Plan (NRRP), Mission 4, Component 2, Investment 1.1, Call for tender No. 104 published on 2.2.2022 by the Italian Ministry of University and Research (MUR), funded by the European Union – NextGenerationEU – Project Title DA VINCI. In Da Vinci’s mind: Fundamental Correlates of Creativity for Artists and Scientists – CUP J53D2300799 0001 - Grant Assignment Decree No. 0001016, adopted on 07.07.2023 by the Italian Ministry of Ministry of University and Research (MUR); the MICS (Made in Italy – Circular and Sustainable) Extended Partnership funded by the European Union Next-GenerationEU, National Recovery and Resilience Plan (NRRP), Mission 4 – Component 2, Investment 2, Investment 1.3 – D.D. 1551.11-10- 2022, PE00000004; and the National Recovery and Resilience Plan (NRRP), Mission 4 - Component 2 - Investment 1.3 RESearch and innovation on future Telecommunications systems and networks, to make Italy more smART (RESTART), project Net4Future, CUP J33C22002880001, PE00000001.
During the preparation of this work, the authors used GPT-4 in order to: grammar and spelling check; and sentence-level proofreading and occasional rephrasing to improve clarity and correctness. All such suggestions were critically reviewed and edited by the authors. After using this tool, the authors reviewed and edited the content as needed and take full responsibility for the publication’s content. doi:10.1093/jcmc/zmac029. [20] P. Kuyer, B. Gordijn, Nudge in perspective: A systematic literature review on the ethical issues with nudging, Rationality and Society 35 (2023) 191–230. URL: https://doi.org/10.1177/10434631231155005. doi:10.1177/10434631231155005. arXiv:https://doi.org/10.1177/10434631231155005. [21] M. Padigar, Y. Li, C. N. Manjunath, “good” and “bad” frictions in customer experience: Conceptual foundations and implications, Psychology & Marketing 42 (2025) 21–43. URL: https:// onlinelibrary.wiley.com/doi/abs/10.1002/mar.22111. doi:https://doi.org/10.1002/mar.22111. arXiv:https://onlinelibrary.wiley.com/doi/pdf/10.1002/mar.22111. [22] M. Cremaschi, M. Dorfmann, A. De Angeli, A steampunk critique of machine learning acceleration, in: Proceedings of the 2024 ACM Designing Interactive Systems Conference, DIS ’24, Association for Computing Machinery, New York, NY, USA, 2024, p. 246–257. URL: https://doi.org/10.1145/ 3643834.3660688. doi:10.1145/3643834.3660688. [23] G. E. Corazza, R. Reiter-Palmon, R. A. Beghetto, T. Lubart, Intelligence and creativity in the space-time continuum for education, business, and development, Journal of Creativity 31 (2021) 100003. URL: https://www.sciencedirect.com/science/article/pii/S2713374521000030. doi:https: //doi.org/10.1016/j.yjoc.2021.100003.