2.1 Identifying current formalisms

To identify current formalisms of design activity, a review of relevant publications in design research was conducted. Specifically, the following journals were reviewed as a starting point: Design Science, Design Studies, Journal of Engineering Design (JED), and International Journal of Design (IJD). These journals represent interdisciplinary design research (Design Science, Design Studies), engineering design focused research (JED), and more industrial design focused research (IJD). Design Studies is the highest impact interdisciplinary design journal, while IJD and JED are the highest rated journals in their sub-domains. These journals thus provide the basis for identifying current formalisms of design activity and the research areas they represent bound the scope of the unification claims reported in this work.

The keywords used for this review were: theory, model, and framework, implemented separately via a full text search, from archive start to 2016. This resulted in 1426 responses for theory, 2007 for model, and 1042 for framework. All responses were filtered by the authors based on whether they explicitly described a ‘formalism’ of design activity. This assessment was based on self-identification by the original authors of the reviewed papers i.e. the original authors specifically describe their formalism as a model, theory or framework. For example, the Design Ontology (Storga et al. Reference Storga2010) was described by Storga et al. as a potential framework for understanding product development work. Similarly, the ontology presented by Sim & Duffy (Reference Sim and Duffy2003) was self-identified as a system model. Where works described formalisms from outside the design domain the original reference was followed up and also included in the review. For example, Beylier et al. (Reference Beylier2009) utilise Markus’ (Reference Markus2001) theory of knowledge reuse. Papers reporting single variable relationships or empirical results only were excluded at this stage. The initial inclusion criteria for the review were thus: (1) published in a major design research journal; (2) explicitly self-identify and describe a formalism (theory, model, framework) of design activity; (3) excluding single variable relationships or purely empirical characterisations. This resulted in the initial identification of 66 unique theories, frameworks, or models (henceforth referred to as formalisms for brevity). Although this list is not exhaustive, sufficient formalisms were reviewed such that common concepts could be robustly identified.

2.2 Analytical approach

A five-step approach was used to analyse the identified formalisms. First, it was necessary to establish at what level prior formalisms describe design activity. This builds on an understanding of human activity as multi-level consisting of activity, task, and action, linked to cognition (Leont’ev Reference Leont’ev1981; Bedny & Harris Reference Bedny and Harris2005). Activity describes a motivation directed aggregation of lower level ‘tasks’, which are themselves aggregations of lower level ‘actions’, which connect directly to human cognition (Bedny & Harris Reference Bedny and Harris2005). Here motivation shapes overall activity, and gives a direction to the system of goals and sub-goals directing tasks and actions respectively (Leont’ev Reference Leont’ev1981). Motivation is long-term, situated, and of varying magnitude. It is possible to hold several motivations at a single time, such as the simultaneous motivations to design a creative product and to build a well-working team (Bedny & Harris Reference Bedny and Harris2005). Lower level goals and sub-goals are thus directed by motivations. Here goals and sub-goals are conscious conceptualisations of a desired future state (Bedny & Harris Reference Bedny and Harris2005). Specifically, the following definitions were applied in the analysis of existing formalisms based on Activity Theory (and are subsequently used throughout for consistency):

Activity level : subjectively distinct periods of behaviour associated with fulfilling a motivation (Leont’ev Reference Leont’ev1981), which are not temporally distinguished. As more than one motivation can be simultaneously held by a designer more than one activity can be progressed at one time, as illustrated by the parallel arrows at this level in figure 1 (Bedny & Harris Reference Bedny and Harris2005). An example activity is creating a novel design concept.

Task level : temporally and subjectively distinct periods of behaviour, associated with fulfilling a goal under specific conditions (Leont’ev Reference Leont’ev1981). Each higher-level activity can link a number of tasks, but the tasks themselves can only occur sequentially (Bedny & Harris Reference Bedny and Harris2005). An example task is brainstorming ideas on paper.

Figure 1. Human activity composed of multiple aggregate levels built on a generic, unitary action/cognition foundation.

Action level : temporally distinct periods of behaviour, associated with achieving immediate, defined sub-goals linked to completing the overarching task (Leont’ev Reference Leont’ev1981; Bedny & Karwowski Reference Bedny and Karwowski2004). Actions form generic building blocks from which higher-level behaviours (tasks and activities) are composed. An example action is representing part of an idea via a sketch.

Cognitive level : a continuous process and structured system of processing units that describes the storage of concepts, propositions, or images etc. as well as the system of internal mental processes underpinning behaviour (Bedny & Harris Reference Bedny and Harris2005). Dominant mental processes also underpin the taxonomic grouping and classification of actions (Bedny & Harris Reference Bedny and Harris2005). An example mental process is embodied cognition.

A ‘slice’ through all four levels provides a description of human activity as the unity of multi-level behaviour and cognition in a given situation (Bedny & Harris Reference Bedny and Harris2005), consistent with both cognitive (Bedny & Karwowski Reference Bedny and Karwowski2004) and behavioural theory (Oliver Reference Oliver1980; Miltenberger Reference Miltenberger2011). This conceptualisation of activity is illustrated in figure 1. Here, higher-level activity can be characterised with respect to unitary building blocks (highlighted in figure 1) at the interface between the action level and cognition level. These unitary building blocks define core actions, which interface directly with cognition and are grouped and taxonomically defined with respect to their dominant cognitive processes (Bedny & Harris Reference Bedny and Harris2005). This has two main implications for evaluation of current formalisms of design activity. First, dominant cognitive processes can be inferred from described behaviours (Scaife & Rogers Reference Scaife and Rogers1996; Wilson Reference Wilson2002) even where such a link is not characterised or described explicitly. Thus all formalisms describing some type of action, task or activity were explicitly included in the presented analysis despite the fact that many of these focus on behavioural rather than unitary (i.e. connecting behaviour and cognition) descriptions. Second, although it is possible to infer cognition connected to behaviour the reverse is not necessarily possible. Formalisms that describe abstract characterisations of the logic underpinning design or cognitive processes that are only generically associated with behaviour cannot be linked to specific actions. For example, Hatchuel & Weil’s (Reference Hatchuel and Weil2002) C–K theory characterises reasoning in design and is only generally connected to behaviour. Thus it cannot be linked to specific actions. Similarly, problem/solution co-evolution (Dorst & Cross Reference Dorst and Cross2001) provides an abstract formalism characterising understanding development in design but again does not connect to specific actions. Formalisms that provide only cognitive or abstract logical descriptions (15 formalisms) were thus excluded from this review because it is not possible to identify the specific unitary ‘action/cognition’ building blocks necessary for a unifying model of design activity. This follows the focus on activity as the unity of behaviour and cognition and introduces a fourth inclusion criterion: formalisms explicitly describe design behaviour or behaviour and cognition in a single model. Thus 51 formalisms were taken forward into the initial analysis (Section 3.1).

In the second step, the identified formalisms from Step 1 were examined in more depth on the action level. When existing formalisms did not explicitly define actions, the articles were further examined to identify the actions underpinning the model. This resulted in a diverse list of specific action sub-types. Formalisms were analysed at the action level for three main reasons. First, actions directly interface with cognition, and can be generically classified in terms of each action’s dominant cognitive process (Bedny & Karwowski Reference Bedny and Karwowski2004). For example, gesture and sketching are distinct behaviours but can both be understood in terms of the representational aspects of external or embodied cognition (Scaife & Rogers Reference Scaife and Rogers1996; Wilson Reference Wilson2002), and thus can be conceptualised as specific instantiations of a generic representation action (Bedny & Karwowski Reference Bedny and Karwowski2004). This allows for a unifying understanding across the diverse descriptions found in the design literature. Second, the action level provides the foundation upon which the higher aggregate levels of task and activity are built (Bedny & Harris Reference Bedny and Harris2005). These higher levels connect strings of actions in any number of sequential configurations dependent on the subject and situation. Finally, actions can be described as a self-regulating system where the individual progresses through a cycle from formulating the goal and assessing its conditions, deciding on and executing the action, and evaluating the outcome to generate a new sub-goal formulation (Engeström Reference Engeström2000). This links to characterisations of behaviour and cognition, critical to understanding causation in this context (Oliver Reference Oliver1980; Miltenberger Reference Miltenberger2011). Thus, it is possible to cohesively connect diverse descriptions of design activity, at the action level via generic actions, consistent with underlying theories of activity, behaviour, and cognition (Leont’ev Reference Leont’ev1981; Miltenberger Reference Miltenberger2011).

In the third step, core actions were identified by clustering the specific action sub-types from Step 2 with respect to the dominant cognitive processes explicitly or implicitly linked to the action in the reviewed formalisms. This assessment was based on the logic that actions provide independently observable and generic building blocks suitable for identifying similarities between formalisms (Engeström Reference Engeström2000; Bedny & Karwowski Reference Bedny and Karwowski2004), and can be grouped and taxonomically classified with respect to their dominant cognitive processes (Bedny & Harris Reference Bedny and Harris2005). Where action sub-types were described purely in terms of behaviour with no explicitly suggested cognitive process, possible dominant cognitive processes were established based on additional literature describing the action sub-type. Thus each action found to be generic and common across formalisms i.e. core, were explored in depth. This enabled us to establish three core actions, which can provide a unifying understanding of design activity.

In the fourth step, connections were identified between the three core actions identified in Step 3. These connections were required because a ‘comprehensive’ model of activity consists of the following elements with respect to the self-regulating action system (Oliver Reference Oliver1980; Engeström Reference Engeström2000; Miltenberger Reference Miltenberger2011): antecedent i.e. the cause or driver for an action, behaviour i.e. the external action itself, and consequence i.e. evaluation of the outcome. The identified core actions provide the behavioural elements and connections between these elements can be established via an antecedent/consequence mechanism. Antecedent/consequence thus forms the specific causal mechanism for the three core actions by defining how they interact with cognition to form a self-regulating system where goal outcomes can be evaluated (Engeström Reference Engeström2000). Such self-regulation typically links basic cognitive (Evans Reference Evans2008) and metacognitive processes (Schraw & Dennison Reference Schraw and Dennison1994). While basic cognitive processes are connected directly to action, metacognitive processes allow an individual to evaluate the limitations of their current knowledge or understanding as well as regulate basic cognitive processes (Schraw & Dennison Reference Schraw and Dennison1994). Metacognitive processes thus form a key bridge from a purely behaviour/cognitive model to one that can account for memory, knowledge, and understanding because they provide knowledge about cognition. They account for individuals’ understanding of their own knowledge state, which forms a key part of design co-evolution (Dorst & Cross Reference Dorst and Cross2001). Metacognition further regulates basic cognitive processes and can thus provide a platform for connecting diverse core actions in a single model (Schraw & Dennison Reference Schraw and Dennison1994). Metacognitive processes have further been demonstrated as important to directing design activity in prior empirical research (Ball & Christensen Reference Ball and Christensen2009). Thus, while each core action must be connected to a dominant cognitive process, a metacognitive antecedent/consequence mechanism is needed to link these core actions.

In the fifth step, the core actions were combined with cognitive and metacognitive processes in a conceptual model to fulfil the research aim. This brings together and connects core actions via common causal mechanisms, consistent with fundamental theories of human activity (Bedny & Harris Reference Bedny and Harris2005; Miltenberger Reference Miltenberger2011) and prior understanding of design activity. Thus, although each element in the model draws from extant literature the final model offers a novel conceptualisation of those elements. This provides a unifying model of design activity and a foundation for future design activity research.

The five steps are reported as follows: Steps 1 and 2 in Section 3.1, Step 3 in Sections 3.2–3.4, Step 4 in Section 4, and Step 5 in Section 5.

3.1 Overview and scope of current formalisms

Table 1. Overview of formalisms associated with design activity, categorised with respect to their related core actions

The 51 identified formalisms are summarised in Table 1, where they are grouped with respect to their constituent core actions. This grouping was not assumed a priori and is for table clarity only. Table 1 outlines the formalisms’ general view on design activity as well as each of the specific action sub-types they describe. Each action sub-type is labelled with respect to its core action, again for clarity: information action (I), knowledge-sharing action (KS), representation action (R). In addition, for each formalism, the cognitive process associated with action is highlighted (labelled as Cog in Table 1). This is either explicit i.e. the original authors state their perspective on the link to cognition with respect to a defined cognitive process (highlighted in bold in Table 1) or implicit i.e. a cognitive process is inferred from the authors’ writing, referencing, and behaviour descriptions because no explicit statement is given. For example, Kavakli & Gero (Reference Kavakli and Gero2001) describe external cognition as the foundation of their work, and cite other papers where this cognitive process has been explicitly highlighted although they themselves do not define an explicit theory of cognition. The 51 identified formalisms were evaluated with regard to the level of their descriptions (based on Figure 1), the characterisations of the action level, and in terms of the processes they describe.

Evaluating the formalisms in terms of level, it was found that current formalisms collectively describe a wide range of specific activities, tasks, and actions. No current formalism describes the full interaction across all levels necessary for understanding design activity completely (Figure 1). For example, one of the most extensive formalisms provided by Sim & Duffy (Reference Sim and Duffy2003) touches on representation but does not deal with sketching, gesture or many of the other representation sub-types. Current formalisms typically focus on the activity or task level, and deal with the designer, their surroundings, underlying cognition, and aggregate activity, but few connect generic actions and their dominant cognitive processes. Further, formalisms offer a large breadth in descriptions of specific action sub-types, ranging from general ‘external or physical expression’ (Hertz Reference Hertz1992; Singh & Gu Reference Singh and Gu2012) to highly detailed breakdowns of representational gesture, sketching, drawing, and writing (Suwa et al. Reference Suwa, Purcell and Gero1998; Sun et al. Reference Sun2014). Descriptions vary in perspectives on the processes underpinning activity and do not deal with core actions able to support wider unification and connection across formalisms. For example, Conversation Theory (Pask Reference Pask1975) focuses on knowledge sharing via language and does not provide for integrating information or representation actions such as those outlined by Vajna et al. (Reference Vajna2005) or Sun et al. (Reference Sun2014). Thus, current formalisms do not provide a basis for unification across levels, based on core actions.

Evaluating existing formalisms on the action level, stark differences were found with regard to described sub-types and their links. No single formalism covered all action sub-types or provided a framework able to synthesise the wide range of descriptions in the literature. This can be partially attributed to the fact that most reviewed formalisms (Table 1) describe design activity disconnected from cognition. Of 51 formalisms only 17 offer some description of the link between action and an explicitly framed cognitive process (Table 1). For example, while the Design Ontology (Storga et al. Reference Storga2010) provides an excellent overview of specific information and knowledge-sharing action sub-types, these are described in behavioural terms without being explicitly linked to a cognitive process. Similarly, Moscoso (Reference Moscoso2007) focuses on actions with respect to the manufacturing process, and thus provides behavioural descriptions without connecting these to cognition. While this general focus on behaviour does not diminish the contribution of the listed formalisms, it does mean that no current formalism is suitable for providing a unifying model of design activity.

Evaluating the formalisms in terms of the processes they describe it was found that most formalisms tend to either describe the antecedent for action or action itself but not both. For example, co-evolution offers detailed insight into the underpinning drivers for action (Dorst & Cross Reference Dorst and Cross2001), but does not connect this to a fully realised framework of activity. Similarly, Sim & Duffy (Reference Sim and Duffy2003) and Storga et al. (Reference Storga2010) detail specific action sub-types but do not tie them to cognitive and metacognitive processes in terms of antecedent/consequence. Thus current characterisation of design activity, built on a self-regulating system connecting core actions and cognition, is incomplete (Oliver Reference Oliver1980; Engeström Reference Engeström2000; Miltenberger Reference Miltenberger2011). This initial analysis results in three distinct limitations preventing the development of a unifying model of design activity based on current formalisms:

1. (1) Lack of cohesive theoretical description of aggregate activity built on core actions linked to cognition.

2. (2) Lack of common and generic core actions able to connect the varied descriptions of design activity currently found in the literature.

3. (3) Incomplete theoretical description from antecedent-to-consequence connecting core actions and cognition.

Due to these limitations, it was necessary to identify core actions that enable unification of design activity. This was done in four phases. First, all explicit cognitive processes were clustered based on common theoretical elements. For example, a cluster of explicit cognitive processes could be identified by their common focus on describing processes where mental structures are externally represented, whether formalised as external cognition, embodied cognition or experiential cognition. This resulted in three clusters: information-based cognition including coordinated information processing (Suss & Thomson Reference Suss and Thomson2012), and information processing (Aurisicchio et al. Reference Aurisicchio, Bracewell and Wallace2013), knowledge-based (group) cognition including memory-based cognition (Liikkanen & Perttula Reference Liikkanen and Perttula2010), group cognition (Pask Reference Pask1975; Dong Reference Dong2005; Dong et al. Reference Dong, Kleinsmann and Deken2013; Eris et al. Reference Eris, Martelaro and Badke-Schaub2014), and collective cognition (Kleinsmann et al. Reference Kleinsmann2012), and representation-based embodied cognition including external cognition (Van Der Lugt Reference Van Der Lugt2005; Bilda & Gero Reference Bilda and Gero2007; Aurisicchio et al. Reference Aurisicchio, Bracewell and Wallace2013), embodied cognition (Daley Reference Daley1982; Oxman Reference Oxman1997; Viswanathan et al. Reference Viswanathan2014), experiential cognition (Demirbaş & Demirkan Reference Demirbaş and Demirkan2003; Gero & Kannengiesser Reference Gero and Kannengiesser2004; Singh & Gu Reference Singh and Gu2012), and thought/language cognition (Fox Reference Fox1981; Eris et al. Reference Eris, Martelaro and Badke-Schaub2014).

Second, the specific action sub-types associated with these three clusters were grouped using explicitly connected formalisms in order to distil core actions characterised by these dominant cognitive processes. For example, the specific action sub-types associated with the initial ‘representation-based embodied cognition’ cluster included formulation, synthesis, analysis, evaluation, documentation, reformulation, language-based representation, physical representation, idea manifestation, external expression, sketching, drawing, gesture, physically storing ideas, physically conveying information, representing ideas, and physically engaging attention (Table 1). Based on this an initial definition was developed for a core action of representation able to accommodate each of these specific action sub-types with respect to a common dominant cognitive process.

Third, this initial core action definition was evaluated with respect to all formalisms that implicitly connected to representation-based embodied cognition, and their associated specific action sub-types. Again these action sub-types were listed and incorporated into the core action conceptualisation where they could be connected to the underlying cognitive process. Where action sub-types did not fit with the underlying cognitive process connected a specific core action, they were compared with other cognitive process clusters. This was the case with, for example, Vajna et al. (Reference Vajna2005), where the action sub-types: evaluation, selection, replication, recombination, and mutation, were all framed within the original text as representation-based actions but were analogous to actions described with respect to formalisms explicitly connected to information-based cognition. This resulted in the identification of a set of action sub-types for each of the three cognitive process clusters.

Fourth, the sets of action sub-types were synthesised through a definition of the core action. These definitions were evaluated iteratively to eliminate possible overlap in action sub-types. This process resulted in three core action definitions based on the reviewed formalisms in Table 1: information action, knowledge-sharing action, and representation action. These core actions bring together the various formalisms from the literature in terms of common cognitive processes.

We propose the three core actions as a foundation for connecting current formalisms of design activity. However, despite the review and distillation of the core actions reported in this section, further refinement is still required due to the lack of definitive descriptions in the existing literature, the high variety of action sub-type descriptions, and relative lack of explicit links to dominant cognitive processes. The review was thus expanded to include the wider design literature where there are many empirical works not described with respect to a specific formalism. For example, Wasiak et al. (Reference Wasiak2010) provide an extensive list of specific action sub-types associated with information action, without explicitly connecting to a formalised model of design activity or to a specific cognitive process. The specific logic underpinning this expanded review is to establish the nature and scope of the core actions and their sub-types in the empirical design literature. This elaboration is necessary to ensure the robustness of the core actions by contrasting and integrating these additional works. Thus the nature of the core actions and their connection with the design and activity literatures via dominant cognitive process and a common metacognitive process is explored in the following sections, which expand the scope of the original review.

3.2 Information action

Information action can be defined as dealing with data parts and their manipulation (Court Reference Court1997). This type of action is associated with cognitive processes describing how data parts are identified, manipulated, and transformed into information and subsequent knowledge (Wilson Reference Wilson1999). This includes the collection, recording, reviewing, and filing of data parts (Blandin & Brown Reference Blandin and Brown1977) via various means including human and non-human (Robinson Reference Robinson2010a ), formal and informal (Fidel & Green Reference Fidel and Green2004). This type of action and its associated cognitive process deal with data parts detached from an individual’s beliefs (Belkin, Oddy & Brooks Reference Belkin, Oddy and Brooks1982; Song, Van Der Bij & Weggeman Reference Song, Van Der Bij and Weggeman2005). It has been shown to be a distinct and important action within design work (Cave & Noble Reference Cave and Noble1986; Puttre Reference Puttre1991), accounting for a significant portion of engineers’ and designers’ time (Court Reference Court1997; Robinson Reference Robinson2010a ). Further, identification and selection of data parts plays a critical role in ideation and design creativity (Gonçalves, Cardoso & Badke-Schaub Reference Gonçalves, Cardoso and Badke-Schaub2016). Some authors even characterise design primarily as an information transformation process where data parts are sought, reasoned about, and stored, via e.g. seeking and requesting (Aurisicchio, Bracewell & Wallace Reference Aurisicchio, Bracewell and Wallace2010; Wasiak et al. Reference Wasiak2010).

Information action has been generally characterised in terms of: finding data parts ${>}$ reasoning about/validating ${>}$ using (Belkin et al. Reference Belkin, Oddy and Brooks1982); driven by perceived need – connected to its role (Belkin et al. Reference Belkin, Oddy and Brooks1982). For example, Wasiak et al. (Reference Wasiak2010) and Cash, Hicks & Culley (Reference Cash, Hicks and Culley2013) characterise information action in the design process in terms of its role e.g. solving. However, research has typically focused on sources and media, such as the internet (Oh, Oh & Shah Reference Oh, Oh and Shah2009), or on aggregate information action with little link to design work, such as total information acquisition (Hult, Ketchen & Slater Reference Hult, Ketchen and Slater2004). Thus although basic taxonomies of information action, such as that by Belkin et al. (Reference Belkin, Oddy and Brooks1982), do exist, and have been included in manifest descriptions such as those offered by Robinson (Reference Robinson2010b ) or Cash et al. (Reference Cash, Hicks and Culley2015), they have not been theoretically integrated with other core actions e.g. representation (Scaife & Rogers Reference Scaife and Rogers1996).

Two main observations emerge from the literature on information action. First, new information is processed through reasoning or learning. It is then structured, and evaluated through experience (Tracey & Hutchinson Reference Tracey and Hutchinson2016) before entering the ‘human belief system’ (Song et al. Reference Song, Van Der Bij and Weggeman2005) to create knowledge. Thus information action must be distinguished from action related to knowledge sharing. Second, information action is connected to designers’ understanding of a situation via their perceived need (Borlund Reference Borlund2003). This has been linked to designer uncertainty perception by Daalhuizen & Badke-Schaub (Reference Daalhuizen and Badke-Schaub2011). Here, designers seek to resolve their uncertainty perception via information action including seeking, gathering, and reasoning about data parts (Kim & Lee Reference Kim and Lee2016). This increases their understanding and thus gradually reduces uncertainty perception over time (Yu et al. Reference Yu2016). The link between uncertainty perception and information action stems back to describing design as a decision making process under uncertainty (Beheshti Reference Beheshti1993) where design progresses as design-related decisions are made and implemented (Wilson Reference Wilson1999; Suss & Thomson Reference Suss and Thomson2012). Thus, uncertainty perception forms a key motivator of information action described in the design literature. Thus, one key mechanism driving information action is the designer’s aim to reduce their uncertainty perception.

3.3 Knowledge-sharing action

Knowledge-sharing action can be defined as dealing with the creation and development of shared understanding (Dong Reference Dong2005). This type of action is associated with cognitive processes describing how knowledge is expressed with respect to an individual’s understanding and beliefs (Court Reference Court1997; Chiu, Hsu & Wang Reference Chiu, Hsu and Wang2006), as well as the structural (i.e. social), relational, and procedural (i.e. shared language or vision) context of an action (Song et al. Reference Song, Van Der Bij and Weggeman2005; Chiu et al. Reference Chiu, Hsu and Wang2006). Thus these actions and their associated cognitive process are fundamentally linked to an individual’s own understanding and beliefs, distinguishing them from information action (Song et al. Reference Song, Van Der Bij and Weggeman2005; Chiu et al. Reference Chiu, Hsu and Wang2006). Knowledge-sharing action forms a critical part of idea sharing (Liikkanen & Perttula Reference Liikkanen and Perttula2010), group creativity (Christensen & Ball Reference Christensen and Ball2016a ,Reference Christensen and Ball b ), group learning (Shull et al. Reference Shull2004), is underpinned by effective communication (Preston, Karahanna & Rowe Reference Preston, Karahanna and Rowe2006), and is often directed towards iteratively grounding shared understanding (Clark & Brennan Reference Clark and Brennan1991) where individuals seek to establish that shared knowledge is actually understood as intended. Thus knowledge-sharing action is characterised by its often discursive nature (Deken et al. Reference Deken2012) where exchanges bring together, fact, rationale, context, varied perspectives, and exploration (Eris Reference Eris2002; Aurisicchio et al. Reference Aurisicchio, Bracewell and Wallace2010). For example, Eris (Reference Eris2002) describes 22 specific types of question, while Aurisicchio et al. (Reference Aurisicchio, Bracewell and Wallace2010) categorise knowledge-sharing requests with respect to their objective e.g. comparison, their subject e.g. product or process, and their response type e.g. retrieval of information. Further, knowledge-sharing action underpins collaborative creative efforts where exchange fosters knowledge acquisition and creation within a group (Shull et al. Reference Shull2004), as highlighted in the recent work of Sauder & Jin (Reference Sauder and Jin2016).

Knowledge-sharing action connects to design formalisms at all levels of description (Maznevski & Chudoba Reference Maznevski and Chudoba2000; Dong Reference Dong2005). However, research has typically focused on understanding knowledge sharing in terms of its role in the development of shared understanding (Preston et al. Reference Preston, Karahanna and Rowe2006; Kleinsmann & Valkenburg Reference Kleinsmann and Valkenburg2008), or its manifestation through various communicative media (Cash & Maier Reference Cash and Maier2016). For example, Maznevski & Chudoba (Reference Maznevski and Chudoba2000) explore team effectiveness in relation to communication intensity; while Aurisicchio et al. (Reference Aurisicchio, Bracewell and Wallace2010) highlight the need to support designers in sharing and capturing knowledge. No single taxonomy of knowledge-sharing action is completely accepted (Eris Reference Eris2002; Aurisicchio, Bracewell & Wallace Reference Aurisicchio, Bracewell and Wallace2006), and despite the fundamentality of knowledge sharing (Fairlie-Clarke & Muller Reference Fairlie-Clarke and Muller2003) it has not been formalised in generic terms or with respect to its relationship with the other core actions.

Two main observations emerge from the literature on knowledge-sharing action. First, it is critically linked to the development of shared understanding (Kleinsmann & Valkenburg Reference Kleinsmann and Valkenburg2008). For example, Conversation Theory (Pask Reference Pask1975) describes the gradual development of alignment in understanding expressed in terms of semantic coherence in word use, as illustrated in design by Dong (Reference Dong2005). Although this is also linked to a number of wider social processes (Busby Reference Busby2001; Chiu et al. Reference Chiu, Hsu and Wang2006), shared context (Humayun & Gang Reference Humayun and Gang2013), and quality of communication (Maznevski & Chudoba Reference Maznevski and Chudoba2000), the development of shared understanding is a core characteristic of successful design teams (Dong Reference Dong2005). It is important to note that although knowledge-sharing action is typically part of an interpersonal exchange it can also be captured in asynchronous modes, where the addressee is unknown or simply imagined, such as in personal letters or journals (Clark & Brennan Reference Clark and Brennan1991; McAlpine, Cash & Hicks Reference McAlpine, Cash and Hicks2017). Thus, although this is a core part of team interaction, the actions themselves are undertaken by the individual. Second, knowledge-sharing action can refer to vision and identity (Chiu et al. Reference Chiu, Hsu and Wang2006), concept understanding (Dong Reference Dong2005), solution understanding (Preston et al. Reference Preston, Karahanna and Rowe2006), and understanding of organisational elements e.g. team roles (Kleinsmann & Valkenburg Reference Kleinsmann and Valkenburg2008). Here, uncertainty perception is particularly connected to others’ understanding of these elements as a driver for knowledge sharing in practice (Clark & Brennan Reference Clark and Brennan1991; Deken et al. Reference Deken2012). Similarly the population of knowledge spaces e.g. problem/solution in co-evolution (Dorst & Cross Reference Dorst and Cross2001), can be related to uncertainty perception via the work of Kreye et al. (Reference Kreye, Goh and Newnes2011) and Christensen & Ball (Reference Christensen and Ball2016a ,Reference Christensen and Ball b ) who connect knowledge sharing, uncertainty perception, and underlying cognitive processes. Here, uncertainty perception is again characterised as a general metacognitive process which drives knowledge-sharing action.

3.4 Representation action

Representation action can be defined as dealing with the perception and manipulation of external representations of information (Scaife & Rogers Reference Scaife and Rogers1996; Wilson Reference Wilson2002). This type of action is associated with cognitive processes describing the interplay between internal and external representations (Scaife & Rogers Reference Scaife and Rogers1996; Wiltschnig et al. Reference Wiltschnig, Christensen and Ball2013). Representation action is often associated with knowledge structures and the exploration of the design space (Dorst & Cross Reference Dorst and Cross2001; Hatchuel & Weil Reference Hatchuel and Weil2003), and provides a cognitively economic means of externalising ideas (Brun, Le Masson & Weil Reference Brun, Le Masson and Weil2016). This has been described via formalisms such as external (Scaife & Rogers Reference Scaife and Rogers1996) and embodied (Wilson Reference Wilson2002) cognition, where an individual uses the interplay between internal/external representations to directly support cognition and develop understanding (Scaife & Rogers Reference Scaife and Rogers1996; Wiltschnig et al. Reference Wiltschnig, Christensen and Ball2013). Representations have also been referred to as simulation in the design literature (Taura et al. Reference Taura2012; Wiltschnig et al. Reference Wiltschnig, Christensen and Ball2013), as such, representation is adopted here for clarity. Specifically, representation actions deal with external representations e.g. prototyping or computational modelling, distinct from internal mental simulation as described by Wiltschnig et al. (Reference Wiltschnig, Christensen and Ball2013). The importance of representation action is highlighted in numerous contexts e.g. via gesture (Cash & Maier Reference Cash and Maier2016), prototyping (Sanders & Stappers Reference Sanders and Stappers2014), or sketching (Schön & Wiggins Reference Schön and Wiggins1992).

Representation action has been widely acknowledged as central to design work (Sim & Duffy Reference Sim and Duffy2003; Horvath Reference Horvath2004; Andreasen et al. Reference Andreasen, Thorp Hansen and Cash2015). In particular, research in this area has focused on understanding and describing representation (Dorst & Vermaas Reference Dorst and Vermaas2005), its relationship to reasoning about design (Dorst & Cross Reference Dorst and Cross2001; Hatchuel & Weil Reference Hatchuel and Weil2003), and how it is realised in practice (Kan, Gero & Tang Reference Kan, Gero, Tang and Gero2011). For example, Kulkarni et al. (Reference Kulkarni, Summers, Vargas-Hernandez and Shah2000) examine how collaborative sketching can support idea generation/manifestation and joint representation in teams. Further, specific aspects of representation action have again been connected to uncertainty perception. For example, Gursoy & Ozkar (Reference Gursoy and Ozkar2015) show how penmanship and sketching help to reveal and resolve uncertainty perception while Scrivener et al. (Reference Scrivener, Ball and Tseng2000) connect sketching and uncertainty perception explicitly in two ways: first, uncertainty perception forms a trigger for strategic shifts within sketching; second, sketching can engender uncertainty perception by elucidating differences between the drawing and the designer’s own understanding and memory. Similarly, Gerber & Carroll (Reference Gerber and Carroll2012) highlight how prototyping also helps to elicit and resolve uncertainty perception, by allowing individuals to iteratively build knowledge and promote a sense of control. However, the centrality of representation action and its interconnection with information action and knowledge-sharing action are not fully captured in current design activity formalisms (Table 1).

Two main observations emerge from the literature on representation action. First, like descriptions of information or knowledge-sharing action, representation is typically described as a multi-faceted phenomenon, linking the design artefact (Gero Reference Gero1990), the specific knowledge being represented (Storga et al. Reference Storga2010), physical modality (Schön & Wiggins Reference Schön and Wiggins1992), and model granularity (Maier, Eckert & Clarkson Reference Maier, Eckert and Clarkson2017). Thus representation action is closely linked to knowledge sharing. For example, representational gesturing can both facilitate individual understanding via external cognition, and the development of team shared understanding through e.g. mirroring and modification (Stempfle & Badke-Schaub Reference Stempfle and Badke-Schaub2002; Cash & Maier Reference Cash and Maier2016). Second, representation is linked to improved understanding and ability to communicate within a team (Schön & Wiggins Reference Schön and Wiggins1992; Scaife & Rogers Reference Scaife and Rogers1996) e.g. by supporting the Concept–Knowledge interaction (Hatchuel & Weil Reference Hatchuel and Weil2003). Here, uncertainty perception is modified via representation surrounding the design artefact as illustrated by the work of Gerber & Carroll (Reference Gerber and Carroll2012) who describe uncertainty perception as a driver for representation action. Thus representation action is again linked to the driver: designer uncertainty perception.

5.1 UDA Elements

The UDA model takes its starting point in the designer’s uncertainty perception . Uncertainty perception forms the antecedent (i.e. causal mechanism) for activity in the UDA model because it motivates the three core actions described below. It connects activity to cognition via the core actions, with respect to memory and experience (Oxman Reference Oxman1990). It forms a metacognitive process that deals with an individuals’ knowledge about their own cognition as well as regulation of cognition itself (Schraw & Dennison Reference Schraw and Dennison1994). It is important to note that although current literature typically highlights the drive to reduce uncertainty perception over time (Ball & Christensen Reference Ball and Christensen2009; Yu et al. Reference Yu2016), UDA uses the more neutral modify. This distinction is made to highlight the fact that actions can both increase or decrease uncertainty perception (Nagai & Gero Reference Nagai and Gero2012). However, overall reduction is to be expected over the whole duration of the design process (Hult et al. Reference Hult, Ketchen and Slater2004). Uncertainty perception can be varied and complex because it includes level and nature, and captures uncertainty perception about the design itself as well as wider organisational issues (Kleinsmann & Valkenburg Reference Kleinsmann and Valkenburg2008). In combination, the level and nature of uncertainty perception can determine action selection and progression in terms of information, knowledge sharing, and representation. Further, changes in any of these dimensions (level, nature or both) form the trigger for shifting between elements within the UDA model. Thus the UDA model allows for the dynamic interaction between changing uncertainty perception and action progression, providing a unifying foundation for describing aggregate activity.

In information action a designer seeks to modify their uncertainty perception by working with data parts and their manipulation. This can be through e.g. collection, recording, reviewing, filing, archiving, seeking, and requesting (Blandin & Brown Reference Blandin and Brown1977; Robinson Reference Robinson2010b ). Individual designers process this via reasoning or learning resulting in a new uncertainty perception state (Daft & Lengel Reference Daft and Lengel1986; Aurisicchio et al. Reference Aurisicchio, Bracewell and Wallace2010). The dominant cognitive process is information processing (Song et al. Reference Song, Van Der Bij and Weggeman2005). Examples of information action include, searching for data online (Aurisicchio et al. Reference Aurisicchio, Bracewell and Wallace2010), or providing/asking for specific data (Robinson Reference Robinson2010a ).

In knowledge-sharing action a designer seeks to modify their uncertainty perception by exchanging knowledge expressed with respect to their understanding and beliefs to e.g. develop a shared understanding with the design team (Daft & Lengel Reference Daft and Lengel1986; Dong Reference Dong2005; Kleinsmann & Valkenburg Reference Kleinsmann and Valkenburg2008). Individual designers process this exchange resulting in a new uncertainty perception state. The dominant cognitive process is social cognition (Chiu et al. Reference Chiu, Hsu and Wang2006). Examples of knowledge-sharing action include, expressing belief modified knowledge in conversation (Dong Reference Dong2005), or creative exploration (Christensen & Ball Reference Christensen and Ball2016a ,Reference Christensen and Ball b ).

In representation action a designer seeks to modify their uncertainty perception by the manipulation of external information representations through, for example, computational simulation (Lamarra & Dunphy Reference Lamarra and Dunphy1998). Individual designers process this by reflecting on the external representation in relation to their internal simulations (Wilson Reference Wilson2002; Evans Reference Evans2008). The dominant cognitive process is embodied cognition (Wilson Reference Wilson2002). Examples of representation action include, sketching (Schön & Wiggins Reference Schön and Wiggins1992), prototyping (Gerber & Carroll Reference Gerber and Carroll2012), or gesturing (Cash & Maier Reference Cash and Maier2016).

In order to describe activity fully, behaviour and cognition must be considered in unity. Thus each of the core actions are connected to cognitive processing (Evans Reference Evans2008), which allows for reflective practice (Schön & Wiggins Reference Schön and Wiggins1992), reasoning, sense-making (Aurisicchio et al. Reference Aurisicchio, Bracewell and Wallace2010), creative imagination (Hatchuel & Weil Reference Hatchuel and Weil2002), and mental simulation (Wiltschnig et al. Reference Wiltschnig, Christensen and Ball2013). Cognitive processing includes both fast and un-deliberate, as well as slow and deliberate processes (Evans & Stanovich Reference Evans and Stanovich2013), however, such decomposition is not necessary for the purposes of the proposed model. In particular, there is a relative lack of unanimity across formalisms of human processing in the psychology literature (Francis et al. Reference Francis2009), as well as their application in the design domain. Thus by combining these elements the UDA model captures the whole cycle from uncertainty perception through behaviour to cognition, as a single unity of activity.

Figure 2. The UDA model linking internal metacognitive uncertainty perception (red) and cognitive processing (grey), and externally enacted information, knowledge-sharing, and representation actions (black). As a whole, the UDA model composes one ‘building block’ denoted by Note 2 in Figure 1.

The UDA model is illustrated in Figure 2. Here the internal world of the designer (including uncertainty perception, cognitive processing, and the reflective link between them) and the external world where action can be observed (including information, knowledge-sharing, and representation actions) are explicitly connected via causal links to produce an overall unity between behaviour and cognition. As such, the model cannot be decomposed to only action or cognition if an overall understanding of activity is to be achieved. Thus, the model must be considered as a whole system. Actions are denoted in black – these include the three core actions as the common generic behaviours from which higher-level tasks and activities are composed; cognitive processing is denoted in grey – this comprises the cognitive processes that underpin the core actions, as well as the other systems making up an individual’s cognitive framework; finally, uncertainty perception is denoted in red – this represents the metacognitive antecedence/consequence mechanism linked to understanding, that connects action and cognition. These elements together make up the unitary action/cognition building blocks at the foundation of activity (Section 2).

All of these elements together make up the foundation of activity, and are couched within the wider context of the situation and the designer (Briggs Reference Briggs2006). A full deconstruction of the different contextual factors that influence activity is beyond the scope of this work and forms a significant body of study in its own right (Bedny & Harris Reference Bedny and Harris2005). The context can include the wider social context (Dong Reference Dong2005), discipline (Yilmaz et al. Reference Yilmaz2015), and personal experience (Christensen & Ball Reference Christensen and Ball2016a ,Reference Christensen and Ball b ). Furthermore, the individual designer links to the wider group (design team and organisation) (Christensen & Ball Reference Christensen and Ball2016a ,Reference Christensen and Ball b ; Kreye Reference Kreye2016). However, all actions in the UDA model are undertaken by an individual (and can all be carried out in any setting, in a group or alone), building on the core understanding of design activity as the unity of individual behaviour and cognition. Group dynamics can be understood by, for example, modelling the actions of each individual and then examining their connection and interaction (Garcia Reference Garcia2005). In this regard it is possible to imagine each individual as semi-autonomous within a network of connected individuals that act independently but influence each other through their actions (McCarthy et al. Reference McCarthy2006). Alternatively, assessment of actions or uncertainty perception can be aggregated across a population in order to identify overall trends or quantitative relationships (Hult et al. Reference Hult, Ketchen and Slater2004). Thus the individual can be connected to the wider group or system by building on the UDA model. However, this interaction is not currently well developed and is substantially beyond the scope of this paper. Finally, deconstructing each element into its sub-constituents e.g. internal processes underpinning cognition, or physical processes and specific instantiations making up the core actions, is also beyond the scope of this work. This is because current formalisms of design activity do not offer consistently detailed and accepted deconstructions of all elements.

5.2 UDA progression

The UDA elements are connected in a unitary model describing the synthesis of behaviour and cognition fundamental to understanding design activity (Bedny & Karwowski Reference Bedny and Karwowski2004), and cyclical action progression (Engeström Reference Engeström2000). Each cycle starts with an antecedent change in the designer’s uncertainty perception. Change in uncertainty perception is used here as a driver for action (Christensen & Ball Reference Christensen and Ball2016a ,Reference Christensen and Ball b ) and can be evaluated in absolute or relative terms (Chan, Paletz & Schunn Reference Chan, Paletz and Schunn2012; Paletz, Chan & Schunn Reference Paletz, Chan and Schunn2017). In reaction to this change in uncertainty perception the designer enters one of the three action cycles. These form the behaviour element of the model. The outcomes from this behaviour then feed into the designer’s cognitive processing system (Bilda & Gero Reference Bilda and Gero2007; Evans Reference Evans2008) where it drives a subsequent change of state in the designer’s uncertainty perception. Cognitive processing (Evans Reference Evans2008) and the change in uncertainty perception form the consequence element in the model and close the cycle. UDA thus captures the full cycle of action from antecedent to consequence, connecting behaviour and cognition (Bedny & Karwowski Reference Bedny and Karwowski2004), to form the foundation for understanding design activity. In addition to the action cycles it is also possible for a designer to simply reflect on their own understanding through a cognitive processing/uncertainty perception loop.

Each cycle forms a generic building block that can be completed multiple times, for example, addressing different information sources (Song et al. Reference Song, Van Der Bij and Weggeman2005) such as searching for specific information on a webpage and then sharing the interpretation of that information with the team, communication modes (Cash & Maier Reference Cash and Maier2016), or representation media (Sanders & Stappers Reference Sanders and Stappers2014). Further, actions can be directed towards different design goals e.g. ideation or concept refinement (Yang Reference Yang2009), or aspects of the design artefact (Gero Reference Gero1990). The core actions and uncertainty perception dynamically interact and co-vary to provide the foundation for describing aggregate activity progression, i.e. the building up of an overall activity by the sequential connection of multiple unitary action cycles each encapsulating behaviour and cognition. Thus UDA is able to describe iterative, dynamic activity via the sequential completion of multiple action cycles.

Figure 3. Possible progressions allowed by the UDA model, and an example sequence of actions linked by evolving uncertainty perception (UP $_{\text{n}}$ ). Here each action cycle in the sequence is a single iteration of the UDA model and together they form the foundation for activity as explained in Figure 1.

The sequential progression through the UDA model is illustrated in Figure 3. This highlights the different possible progressions allowed by the model, as well as examples of the features of design work that these describe. Each cycle is initiated by the current uncertainty perception state (UP $_{1}$ ) and ends with a new uncertainty perception state (UP $_{2}$ ). Finally, a simplified abstract illustration is used to show the dynamic interaction between changing uncertainty perception and action progression.

Together the proposed elements and cyclical progression allow for a model that offers an integrative understanding of design activity built on a foundation of the three core actions connected to cognition. This is consistent with current design activity formalisms. Further, by integrating these elements the UDA model is able to describe the major features of design activity found in the empirical literature.