1. Introduction
To tackle the growingly complex social-technical challenges facing humanity, the design process demands higher-order and higher-dimensional creativity. At the same time, the rapid development of Artificial Intelligence (AI)—particularly generative AI—offers new opportunities to foster Artificial Creativity (AC) in design. AC aims to simulate human creative behavior through computational systems. It has evolved from early frameworks that defined creativity as the exploration of possibilities (Reference BodenBoden, 2004) to recent approaches incorporating generative AI techniques (Reference Berns and ColtonWyse, 2019; Reference WyseBerns & Colton, 2020).
The term Artificial Creativity was first introduced to the design field by Saunders and Gero (Reference Saunders and Gero2001). However, key challenges in AC research remain. These include the need for simultaneous development of design problems and solutions to reflect the inherent complexity of creative design (Reference CrossCross, 1999; Saunders & Gero, Reference Saunders and Gero2001), as well as the challenge of balancing novelty and effectiveness to ensure that generated concepts are both innovative and practically applicable. By integrating human-centered perspectives and expanding the opportunity space, AC systems have the potential to significantly enhance designers' creative processes.
Despite the emergence of various AI-driven design tools, a theory-based framework for implementing artificial creativity in design is still lacking. This paper aims to address this gap by proposing an AC framework grounded in creative cognition and design theories. This framework can serve as a guide for developing and evaluating AI-enabled tools that augment design creativity. By offering a structured approach, the framework can help ensure that AC systems not only generate ideas but also contribute meaningfully to real-world design challenges.
2. Background literature
2.1. Design problem and solution spaces
Creative design involves the dynamic exploration of both problem and solution spaces (Reference CrossCross, 1999). The problem space focuses on understanding how humans interact with their environments, encompassing physical and mental levels (Reference Visser, Stappers, Lugt and SandersVisser et al., 2005; Surma-Aho & Hölttä-Otto, Reference Surma-Aho and Hölttä-Otto2022). The solution space, in contrast, centers on product features and functions, particularly innovative attributes that differentiate new designs from existing solutions (Reference ArthurArthur, 2007). The interplay between these spaces is essential for generating creative outcomes that effectively address human needs.
To explore the problem space, designers utilize methods such as contextual inquiry (Reference Dekker, Nyce and HoffmanDekker, 2003; Karen & Sandra, Reference Karen and Sandra2017), diary studies (Reference Sohn, Li, Griswold and HollanSohn et al., 2008), and participatory design (Reference Visser, Stappers, Lugt and SandersVisser et al., 2005), which directly involve users in the design process. Moreover, human-centered design frameworks like positive design extend the scope of the problem space by focusing on not only user needs and requirements but also on enhancing psychological well-being through experiences (Reference Desmet and PohlmeyerDesmet & Pohlmeyer, 2013; Desmet & Fokkinga, Reference Desmet and Fokkinga2020).
In the solution space, designers utilize heuristics such as TRIZ (Reference YonelinasAltshuller, 2002), knowledge-based systems (Reference Luo, Sarica and WoodLuo et al., 2021), and function modeling (Reference Sridharan and CampbellSridharan & Campbell, 2005) to generate novel solutions by expanding access to technical knowledge and functional alternatives. Cross-domain knowledge synthesis, such as biologically inspired design (BID), draws on principles from nature to inform novel design concept generation (Reference Fayemi, Maranzana, Aoussat and BersanoFayemi et al., 2014).
The iterative co-development of problem and solution spaces promotes a dynamic and integrated design process. Unlike linear approaches, this iterative process allows evolving design problems to align closely with innovative solutions (Reference CrossCross, 1999; Kimbell, Reference Kimbell2014). By simultaneously reframing problems and generating solutions, designers can uncover new opportunities for creativity and innovation.
2.2. Cognitive processes and dimensions of creativity in design
Creativity in design emerges through cognitive processes involving both semantic and episodic memories. Semantic memory provides general knowledge, while episodic memory draws on personal experiences, both of which are crucial for creative ideation (Reference TulvingTulving, 2002; Reference Benedek, Beaty, Schacter and KenettBenedek et al., 2023). The interplay between these types of memory supports associative processes and enables designers to recombine experiences into novel contexts, facilitating the generation of new ideas (Reference Schacter and AddisKenett & Faust, 2019; Schacter et al., Reference Kenett and Faust2007). This interplay allows designers to explore new associations and envision future scenarios beyond their personal experiences (Reference Yin and ChildsYin & Childs, 2024).
Benedek et al. (Reference Benedek, Beaty, Schacter and Kenett2023) integrates insights from cognitive psychology into a framework of creativity consisting of four stages: memory search (i.e., retrieving relevant knowledge and experiences), idea construction (i.e., forming initial creative concepts through associations), novelty evaluation (i.e., assessing the originality of ideas), and effectiveness evaluation (i.e., determining the feasibility and usefulness of ideas), as summarized in Figure 1. The framework emphasizes how controlled memory retrieval, dynamic associations, and constructive recombination foster creativity.
Memory processes in creative ideation (Reference Benedek, Beaty, Schacter and Kenettadapted from Benedek et al., 2023)
Creativity in design extends beyond intrinsically motivated ideation, which involves the episodic memory process of recalling personal experiences; it requires empathy to understand users' experiences deeply (Reference Grant and BerrySaunders & Gero, 2001; Grant & Berry, Reference Saunders and Gero2011). Empathy increases creativity by allowing designers to adopt the user's perspective, thereby broadening the design problem space and uncovering new design opportunities (Reference Alzayed, Miller and McCombRaviselvam et al., 2016; Reference Raviselvam, HölttäOtto and WoodAlzayed et al., 2021). This empathic capacity is critical for developing effective human-centered design concepts.
Creativity can be assessed based on two key dimensions: novelty and effectiveness (Reference Benedek, Beaty, Schacter and KenettBenedek et al., 2023). Novelty involves challenging established norms or developing unexpected solutions (Reference Plucker, Beghetto and DowPlucker et al., 2004; Reference Dean, Hender, Rodgers and SantanenDean et al., 2006). Novelty can be further divided into solution novelty—the development of new technologies, functions, or design features—and problem novelty—redefining user needs and reframing problems to uncover innovation opportunities (Reference Jansen, Van Den Bosch and VolberdaJansen et al., 2006). Effectiveness in design is characterized by two key aspects: feasibility, the practical viability of implementing a solution given technical capabilities, resources, and constraints (Reference Dean, Hender, Rodgers and SantanenDean et al., 2006), and usefulness, which refers to meeting user needs and creating value (Reference Plucker, Beghetto and DowPlucker et al., 2004; Grant & Berry, Reference Grant and Berry2011). Effectiveness balances originality with practicality, ensuring designs are both innovative and applicable. Together, these dimensions—novelty and effectiveness—provide a robust framework for evaluating creativity in design (Figure 2), ensuring that outcomes are both original and impactful.
Dimensions of creativity in design
2.3. From artificial intelligence to artificial creativity in design
Artificial Intelligence (AI) has become a powerful tool in supporting early-stage design processes, particularly in concept generation and user research. AI-based and data-driven innovation (Reference LuoLuo, 2022) expand the opportunity and design spaces for designers by discovering and evaluating design opportunities and generating and evaluating solution concepts based by leveraging large datasets. Knowledge-based systems aid designers by retrieving and synthesizing cross-domain knowledge, enabling out-of-box thinking during ideation (Reference Luo, Sarica and WoodLuo et al., 2021; Reference Vattam, Wiltgen, Helms, Goel and YenVattam et al., 2011; Deldin et al., Reference Deldin, Schuknecht, Goel, McAdams and Stone2014).
Generative AI techniques, such as large language models (LLMs) and Generative Adversarial Networks (GANs), have further pushed the boundaries of AI's potential in design. For instance, Zhu and Luo (Reference Zhu, Zhang and Luo2023) demonstrated that LLMs can automate functional synthesis and biologically inspired design, while Yuan et al. (Reference Yuan, Marion and Moghaddam2023) used GANs to create visual representation of design concepts that align with user desirability. These advancements show that generative AI can enhance the creative capacities of designers by expanding both the solution space and the problem space.
AI has also proven valuable in user research for exploring the problem space. Techniques such as natural language processing (NLP) may extract insights from user-generated content, revealing preferences, sentiments, and experiences (Reference Wang, Lu and TanWang et al., 2018; Siddharth et al., Reference Siddharth, Blessing and Luo2022b). However, traditional AI approaches tend to focus on exploitative processes and emphasize existing features and needs rather than uncovering novel opportunities. To address this limitation, recent developments in generative AI have introduced methods for empathic understanding. For example, LLMs can infer implicit user motivations, providing a deeper and more comprehensive understanding of human needs (Reference Zhu, Chong, Yang and LuoZhu et al., 2025). These capabilities enable AI to uncover novel design problems and opportunities, broadening the scope of the problem space.
In summary, AI-driven tools and methods are evolving from supporting intelligence to enabling Artificial Creativity (AC) — systems that simulate human creative behaviors through computational means. By combining human-centered perspectives with AI’s ability to expand opportunity and design spaces, AC systems have the potential to significantly enhance the design process. These systems go beyond generating ideas; they co-create with designers, helping to discover new problems, develop innovative solutions, and ultimately push the boundaries of creative design. In the following, we present a theoretically-grounded framework for guiding the development of AC systems.
3. A new framework of artificial creativity for design
Our proposed Artificial Creativity (AC) framework integrates principles from design cognition theories and AI methodologies. The framework as illustrated in Figure 3. emphasizes the co-development of design problem and solution spaces, a critical factor for fostering creativity in complex design tasks.
3.1. Spaces in the AC framework
Design research, particularly C-K theory, distinguishes between knowledge and concept spaces during the creative design process (Reference Kazakçi and TsoukiàsKazakçi and Tsoukias, 2005). Knowledge spaces consist of validated knowledge, while concept spaces include speculated ideas that cannot yet be proven or disproven (Reference Kazakçi and TsoukiàsKazakçi and Tsoukias, 2005). Inspired by Kimbell (Reference Kimbell2014), we integrate these concepts into a unified framework, identifying four spaces for creative design: Solution-Knowledge space (KS), Solution-Concept space (CS), Problem-Knowledge space (KP), and Problem-Concept space (CP).
The design problem-solution co-development framework of AC
Solution-Knowledge space (KS): This space contains the validated knowledge relevant to developing functional design solutions. It includes technical data, proven design knowledge, and cross-domain knowledge. The KS space ensures that concepts generated by the AC system are technically feasible and grounded in established knowledge.
Solution-Concept space (CS): This space involves speculative and evolving solution ideas that have not yet been validated. It contains potential product functions and structures, offering possibilities for innovative configurations and features.
Problem-Knowledge space (KP): This space holds established knowledge about design problems, including user needs, preferences, mental states, usage contexts, and market insights. The KP space is informed by extensive user research and serves as the foundation for framing meaningful design challenges.
Problem-Concept space (CP): This space focuses on emerging definitions of novel design problems. This space involves rethinking and reframing the design problem based on in-depth user insights or envisioning opportunities of designing for future user experiences, allowing for the discovery of latent and novel design problems. Unlike the KP space, the CP space emphasizes the imaginative aspects problem discovery.
The design of Dyson's first bagless vacuum cleaner exemplify the four spaces. This product was the first to address the need to eliminate dust bags from vacuum cleaners by introducing cyclone technology from industrial equipment—a compelling example of a novel solution to a novel problem. In this case, the KS entity is “the cyclone technology that uses centrifugal force to separate dust from air.” The CS entity can be described as “integrating a cyclone into a vacuum cleaner to collect dust.” The KP entity is “the users' need for convenience in replacing dust bags.” The CP entity that envisions a new design opportunity is “a mechanism that collects dust, including a reusable container that can be easily cleaned after use.”
3.2. Operators in the AC framework
To automate creative processes within these spaces, we introduce nine AC operators, grouped into forward operators and backward operators. Figure 4 is a visual representation of how the operators function in and across spaces. These AC operators simulate cognitive processes to expand, associate, and evaluate concepts dynamically.
3.2.1. Forward operators
The forward operators include retrieval, synthesis, inference, envision and mapping. They expand the four spaces through simulating various cognitive processes to create an extensive divergence of possibilities where new concepts of problems and solutions could arise. The forward operators include:
Retrieval: The retrieval operator expands the KS space by retrieving knowledge from project databases, knowledge-based systems, cross-domain knowledge sources, or pre-trained models (Reference Luo, Sarica and WoodLuo et al., 2021; Sarica et al., Reference Sarica, Song, Luo and Wood2021; Zhu et al., Reference Zhu, Zhang and Luo2023; Siddharth & Luo, Reference Siddharth and Luo2024). The aim is to construct a KS space that expand designers' semantic memory, encompassing both base-field knowledge directly associated with the project and far-field, cross-disciplinary insights that inspire design and foster new concept properties. For example, the Dyson vacuum cleaner design integrated the “industrial cyclone that uses centrifugal force to separate dust from air”. Industrial cyclone was a far-field knowledge to “vacuum cleaner”.
Synthesis: The synthesis operator expands the CS space by integrating elements from the KS space to generate new solution concepts. This involves constructing semantic links between knowledge entities (Reference BodenBoden, 2004; Reference Benedek, Beaty, Schacter and KenettBenedek et al., 2023; Yin & Childs, Reference Yin and Childs2024) using creative design heuristics that enable the synthesis and transfer of knowledge into properties of solution concepts. Heuristic reasonings like TRIZ, combinational creativity, and design-by-analogy facilitate this process and have been automated by LLMs recently (Jiang & Luo, (Reference Jiang and Luo2024; Reference Zhu and LuoZhu & Luo, 2023; Reference Jiang, Hu, Wood and LuoJiang et al., 2022).
Inference: The inference operator expands the KP space by deriving implicit insights from user data, such as interview transcripts, ethnographic data, user observation records, and large-scale user-generated content to infer underlying emotions, needs, and motivations of users. Achieving this requires advanced cognitive abilities such as perspective-taking and mental simulation to foster a deep, empathic understanding of users’ experiences (Reference GaesserGaesser, 2013; Surma-Aho & Hölttä-Otto, Reference Surma-Aho and Hölttä-Otto2022). AI techniques like NLP significantly enhance this process. For example, LLMs have been used to draw empathic mental inference for understanding users’ underlying goals and psychological needs (Reference Zhu, Chong, Yang and LuoZhu et al., 2025).
Envision: The envision operator expands the CP space by reflecting on existing experiences within the KP space and imagining potential future experiences of people. This helps avoid the “empathy trap” (Reference Mattelmäki, Vaajakallio and KoskinenMattelmäki et al., 2014), i.e., fixation on familiar problem framing, and promotes divergent thinking. The process requires designers to construct detailed mental images of future experiences through complex episodic simulation (Reference Schacter and AddisSchacter et al., 2007; Reference Addis, Pan, Musicaro and SchacterAddis et al., 2016). While various human-centered design methods have been explored to facilitate envisioning future alternatives (Reference Dekker, Nyce and HoffmanDekker et al., 2003; Mattelmäki et al. Reference Mattelmäki, Vaajakallio and Koskinen2014; Desmet & Pohlmeyer, Reference Raviselvam, HölttäOtto and Wood2013; Raviselvam et al., Reference Desmet and Pohlmeyer2016), the computational implementation of the envision operator remains unexplored in the literature. Future research could combine analytical and generative models to address this gap.
Mapping: The mapping operator associate concept properties in and across the CS and CP spaces, facilitating the organization of ideas and identification of promising design concepts. This mirrors how designers associate and organize insights generated during brainstorming to support informed decision making (Reference Holtzblatt and BeyerHoltzblatt & Beyer, 1993). Techniques like unsupervised clustering techniques and supervised graph neural networks (GNN) can be employed to organize concepts, uncover patterns, and predict connections among concept properties. The mapping process could be further supported by employing LLMs for qualitative analysis, enabling the derivation of patterns and structures that underlie the concept space (Reference Xiao, Yuan, Liao, Abdelghani and OudeyerXiao et al., 2023).
Forward operators are responsible for expanding the design problem and solution spaces by simulating cognitive processes. These operators enable the generation of diverse possibilities by retrieving relevant knowledge, combining ideas, envisioning future opportunities, and mapping relationships within and across the Solution-Knowledge (KS), Solution-Concept (CS), Problem-Knowledge (Kp), and Problem-Concept (CP) spaces.
A visual illustration of how operators function in and across spaces
3.2.2. Backward operators
The backward operators focus on evaluating concepts and iterating through the creative design process based on the four dimensions identified in Section 2.2. They help navigate the exploration within and across the four spaces to converge on desirable concepts. The backward operators include:
Solution novelty: The solution novelty operator evaluates the originality of generated solutions by analyzing their divergence from existing knowledge within the KS space. This involves assessing how technical and functional features of a solution differ from those of known alternatives. Grounded in cognitive psychology, novelty evaluation compares new ideas with relevant knowledge stored in long-term memory (Reference Benedek, Beaty, Schacter and KenettBenedek et al., 2023; Yonelinas, Reference Yonelinas2002). Methods such as semantic distance leverage text embeddings to quantify the uniqueness of ideas, providing a measurable basis for novelty assessment (Reference Camburn, He, Raviselvam, Luo and WoodCamburn et al., 2020; Zhu & Luo, Reference Zhu and Luo2023). Recent advancements in LLMs further enable automated evaluations of solution novelty, offering scores that reflect the degree of divergence from existing knowledge (Reference Huang, Huang, Liu, Luo and LuHuang et al., 2025).
Problem novelty: The problem novelty operator focuses on the originality of newly framed design problems by comparing them to established knowledge within the KP space. It identifies the extent to which a new problem definition departs from conventional understandings and aligns with emerging opportunities. Like solution novelty, problem novelty relies on semantic comparisons to evaluate distinctiveness, with techniques such as semantic distance providing a structured approach to measure divergence (Reference Beaty and JohnsonBeaty & Johnson, 2021; Reference Beaty, Johnson, Zeitlen and ForthmannBeaty et al., 2022). LLMs further enhance this process by identifying unconventional problem definitions (Reference Organisciak, Acar, Dumas and BerthiaumeOrganisciak et al., 2023), enabling designers to explore more creative and impactful problem spaces.
Concept feasibility: The concept feasibility operator assesses the practicality of implementing a solution, considering technical constraints, resources, and current capabilities within the KS space. Semantic memory helps designers assess feasibility by recalling similar solutions and their practical implementations (Reference Benedek, Beaty, Schacter and KenettBenedek et al., 2023). Surveys and expert evaluations traditionally measure feasibility, but AI techniques can enhance this process by automating the retrieval of feasibility-related knowledge. LLM-driven Design Structure Matrix (DSM) can facilitate the identification of potential design conflicts and dependencies (Reference Jiang, Xie and LuoJiang et al., 2024), while Retrieval-Augmented Generation (RAG) approaches, which combine knowledge retrieval and generative capabilities (Reference Lewis, Perez, Piktus, Petroni, Karpukhin, Goyal and KielaLewis et al., 2020), can also support feasibility assessments by automating the retrieval of similar engineering solutions to assess whether the concept was achievable with available resources. Such technique can integrate diverse knowledge sources and assist designers in evaluating radical designs that require a broader span of semantic memory.
Concept usefulness: The concept usefulness operator evaluates the potential value of a design concept in meeting user needs and providing tangible benefits, drawing on insights from the KP space. While current research supports AI-driven prediction of user satisfaction for existing product features (Reference Wang, Lu and TanWang et al., 2018), imagining the unfamiliar experiences of new concepts remains a challenge (Reference NormanNorman, 2004). Episodic memory and mental simulation play crucial roles in envisioning the user experience of new designs (Reference Nielsen, Escalas and HoefflerNielsen et al., 2018; Reference Benedek, Beaty, Schacter and KenettBenedek et al., 2023). LLMs can assist by simulating user scenarios and predicting how well a concept integrates into user activities and addresses their needs. This capability leverages a more holistic understanding of user contexts and experiences to form evaluation criteria and elaborate on concept properties. This enhances the evaluation of usefulness during the conceptual stage and ensures that design concepts are more closely aligned with user needs.
Evaluating both feasibility and usefulness is inherently complex due to insufficient understanding of the cognitive mechanisms involved (Reference Benedek, Beaty, Schacter and KenettBenedek et al., 2023), leading to a significant risk of underestimating the potential of radical designs with novel features or applications. For instance, the heavier-than-air flight machine was initially deemed infeasible, yet it revolutionized transportation once proven viable. Similarly, early perceptions of cell phones framed them as useful only for business professionals, but they ultimately became indispensable to almost everyone (Reference NormanNorman, 2004). These examples highlight the challenge of assessing concepts that demand extended memory spans and sophisticated cognitive processes (Reference Nielsen, Escalas and HoefflerNielsen et al., 2018). Recent advancements in RAG and LLMs offer promising opportunities to activate relevant knowledge within expansive knowledge spaces of KS and KP, enabling a more comprehensive and informed evaluation process. Overall, backward operators ensure that the design process remains iterative and grounded in meaningful evaluation. By assessing novelty (both solution and problem) and effectiveness (feasibility and usefulness), these operators help refine concepts, ensuring that final outputs are innovative, practical, and aligned with user needs.
3.3. Human-AC co-design potential
In the context of Human-AC co-design, the creative design process can be broadly categorized as either problem-driven or solution-driven, with designers playing a pivotal role in guiding and shaping the process. A problem-driven strategy focuses on addressing specific needs and their root causes. Designers initiate this process by supplying initial knowledge related to the Problem-Knowledge (KP) and Solution-Knowledge (KS) spaces. Conversely, a solution-driven strategy begins with technical knowledge or conceptual solutions and explores their novel applications. When problem knowledge is absent, the AC system can automatically gather stakeholder data, such as user-generated content or past interview records, to construct an initial problem space for exploration.
Designers actively refine the process by exploring the knowledge spaces (KP and KS) to identify new insights and inspire additional concept properties. This exploration helps expand the opportunity space and uncover novel design possibilities. At the same time, designers can impose constraints on the design spaces to guide the direction of the exploration. When constraints are applied to the problem space, the process becomes problem-driven, focusing on addressing specific design challenges as they arise. On the other hand, constraints on the solution space steer the process toward a solution-driven approach, emphasizing the discovery of new applications for existing technologies or concepts.
This dynamic interplay between exploration and constraint transforms AI from a mere design tool into an active co-creator. The AC system collaborates with designers to explore, redefine, and expand the boundaries of creativity. By facilitating the simultaneous development of design problems and solutions, Human-AC co-design enables the discovery of innovative opportunities that might otherwise remain hidden. This collaboration leverages the strengths of both human intuition and AI’s capacity to process and generate vast amounts of data, resulting in a more comprehensive and creative design process.
Ultimately, Human-AC co-design enhances the ability to tackle complex design challenges by combining human-centered insights with AI’s expansive potential. This approach fosters breakthrough innovations by enabling designers and AI to work together to discover new problems, envision novel solutions, and push beyond conventional design boundaries. In doing so, Human-AC co-design represents a significant step forward in augmenting human creativity with the power of Artificial Creativity systems.
4. Discussion
In this paper, we propose an Artificial Creativity (AC) framework aimed at facilitating the simultaneous exploration of design problems and solutions. The framework employs a series of operators to expand the knowledge space, transform knowledge into divergent conceptual properties, and help converge on concepts that are both novel and effective. Each operator aims to operationalizes specific type of design theories and methods, some of which have been automated through existing AI technologies, while others remain underexplored.
Historically, design theories and methods have often been viewed as abstract and subjective, lacking clear rules and pathways for automation. As a result, the design processes have typically relied on the cognitive capabilities of human designers. However, recent advancements in LLMs have demonstrated their ability to simulate human cognitive processes and perform various cognitive tasks with remarkable proficiency (Reference Binz and SchulzBinz & Schulz, 2023; Reference Demszky, Yang, Yeager, Bryan, Clapper, Chandhok and PennebakerDemszky et al., 2023). Some of these cognitive capabilities, such as knowledge retrieval, analogy-making, and empathic inference, have already shown promise in design applications (Reference Zhu and LuoZhu & Luo, 2023; Reference Zhu, Chong, Yang and LuoZhu et al., 2025).
Generative AI, particularly LLMs, offers significant potential for the computational implementation of various design theories and methods—particularly those whose cognitive reasoning processes were previously accessible only to human designers. This opens new avenues to augment and enhance the creative design process in ways that were previously unattainable. By grounding these efforts in a robust theoretical framework, we can effectively synthesize AI capabilities with design cognition principles, providing a more structured approach to developing AI-enabled design tools.
The problem space considered in this paper focuses primarily on human-centered perspectives, addressing issues or opportunities that arise from user experiences. However, design also has the potential to influence broader ecosystems and address long-term sustainability challenges (Reference NormanNorman, 2023). Expanding the problem space to include these broader, systemic issues could lead to impactful solutions that benefit not only individuals but also communities and the environment. Incorporating such an expanded view into the AC framework will require operators capable of reasoning about complex, interrelated factors affecting both people and ecosystems, which lies beyond the scope of this paper.
Future research should explore how AC systems can effectively address these broader challenges. This may involve developing new operators that integrate systems thinking, sustainability considerations, and ethical design principles. Additionally, further investigation is needed to refine the evaluation mechanisms for concept feasibility and usefulness, especially for radical and forward-thinking design solutions. While current methods rely heavily on existing knowledge, breakthroughs often require envisioning possibilities that go beyond current constraints and assumptions.
In conclusion, our proposed AC framework provides a theoretically grounded approach to augmenting design creativity with AI. By balancing novelty and effectiveness, and by facilitating the co-development of problems and solutions, this framework can help unlock new creative potentials. As AI technologies continue to advance, integrating these capabilities with human-centered design principles will be essential for addressing the increasingly complex socio-technical challenges of our time.
Open access
