Complex problem solving: a case for complex cognition?

Complex problem solving (CPS) emerged in the last 30 years in Europe as a new part of the psychology of thinking and problem solving. This paper introduces into the field and provides a personal view. Also, related concepts like macrocognition or operative intelligence will be explained in this context. Two examples for the assessment of CPS, Tailorshop and MicroDYN, are presented to illustrate the concept by means of their measurement devices. Also, the relation of complex cognition and emotion in the CPS context is discussed. The question if CPS requires complex cognition is answered with a tentative “yes.”


Introduction
In the last 30 years, a new part of the psychology of thinking and problem solving emerged in Europe under the label of complex problem solving (CPS).This paper introduces into this field and provides a review of related concepts from my personal perspective.A major goal of this selective review is a presentation of research findings and specific (German) research traditions that are not so much recognized internationally up to now (for an exception, see Osman 2009).
Central for this paper is an analysis of the question if CPS requires complex cognition.Therefore, I first try to define complex cognition; second, I introduce conceptually related concepts like macrocognition or operative intelligence to show the overlap of the concepts as well as their specific contributions.Third, the assessment of CPS is illustrated by two measurement approaches, Tailorshop and MicroDYN.Fourth, I point out the interaction between cognition and emotion in complex situations.Finally, some conceptual problems with the term ''complex cognition'' are discussed and an answer to the question in the title will be given.

Complex cognition
According to Knauff and Wolf (2010), complex cognition deals with all mental processes that are used by an individual for deriving new information out of given information, with the intention to make decisions, solve problems, and plan actions.This approach assumes an active and goal-directed information processing by human beings who are able to perceive their environment and to use their memory.The term ''mental processes'' is not restricted to cognitive ones but includes motivational and emotional processes at the same time.
In this understanding, complex cognition can be contrasted to simple cognition: the latter refers to the study of elementary cognitive processes like perception, memory, or learning in isolation and without asking for broader context.But are simple and complex cognition two dichotomous categories or rather two extremes of a complexity continuum?Is there a line of demarcation between them or is the complex case simply an extension of elementary processes?
An example of simple cognition is the analysis of color perception-for example, the perception of a red light-under conditions of reduced brightness.In contrast, complex cognition could be found in a case where a ship's captain oversaw a red warning light in the dark and is now confronted with the immediate task of accident prevention.
Whereas in the first case, a theory of color perception might help to predict the circumstances under which a red light might be overseen, in the second case, a theory of complex cognition would put the issue of an overseen red light in a broader context and might predict short-term consequences like stress, shame, and guilt processes, together with activities for error correction, and long-term consequences like increased attention to warning lights in general.The first case illustrates a cognitive process of a ''pure'' isolated psychic function without a meaningful action goal; the second case shows the orchestration of the whole set of cognitive functions including, for example, emotions and serves an overarching goal of action regulation like accident prevention.
Is the complex process completely understandable in terms of the addition of simple elements?Yes and noyes, for sure, simple processes are involved in the complex situation; no, because the result of this summation follows the Gestalt principle, whereas the whole is more than the sum of its parts (von Ehrenfels 1890).In the complex situation, the result is more than the sum of perceptual, learning, and memory processes-it is an organized, hopefully well-orchestrated stream of actions serving goals of an acting individual.It is not perception or memory on its own-it is perception as part of a higher structure; it is embedded into and tuned to other psychic processes, which serve action goals.The context delivers the meaning: in our example from above, the (neutral) color red becomes a warning sign in the context of a control room, its detectability and meaning are not only given by itself but in combination with other signs and signals.The contextual embedding allows also to integrate cultural effects (Nisbett and Norenzayan 2002) and, thus, transcends a simple addition of the underlying microprocesses.
The move from the level of elementary processes to complex cognition is a shift in the hierarchy level of explanation (see for more details Rasmussen 1983): instead of talking about the detectability of red lights as a signal (level 1), one is talking about the red light as a stop sign (level 2) or even further about the stop sign as part of a broader concept of traffic regulation (level 3).The predictions (and the corresponding assessments) would be different according to these three types of ''situation awareness'' (Endsley 1995a, b).The line of demarcation is a matter of different levels: staying on the lowest level means simple cognition, going up the hierarchy level implies increasingly more complex cognition.

What is a complex problem?
In current theories of problem solving, a problem is conceptualized as composed of a given state, a desired goal state, and obstacles between given and goal state (Mayer 1992, p. 5).According to Anderson (2005), ''any goaldirected sequence of cognitive operations'' is classified as an instance of problem solving.Other authors supplement this view by emphasizing that overcoming barriers toward a desired state may involve both cognitive and behavioral means, and that these means should imply novelty and originality.Operations triggered by a problem go beyond routine action and thinking (Frensch and Funke 1995a, b).
A complex problem is said to occur when finding the solution demands a series of operations which can be characterized as follows (Do ¨rner et al. 1983): Elements relevant to the solution process are large (complexity), highly interconnected (connectivity), and dynamically changing over time (dynamics).Neither structure nor dynamics are disclosed (intransparency).Finally, the goal structure is not as straight forward as suggested above: in dealing with a complex problem, a person is confronted with a number of different goal facets to be weighted and coordinated-a polytelic situation.
One approach to complex situations falls under the label of ''Naturalistic Decision-Making'' (Klein 2008;Klein et al. 1993;Lipshitz et al. 2001;Zsambok and Klein 1997).According to Zsambok (1997, p. 5), ''the study of NDM asks how experienced people working as individuals or groups in dynamic, uncertain, and often fast-paced environments, identify and assess their situation, make decisions, and take actions whose consequences are meaningful to them and to the larger organization in which they operate.''The four markers of this research approach-(a) complex task in real-life setting, (b) experienced decisionmakers, (c) actual decision-making, and (d) situation awareness within an decision episode-are quite different to the approach frequently chosen in psychology using artificial decision situations with a focus on option selection, student subjects, and a rational standard (e.g., Kahneman et al. 1982).
Another approach falls under the label of ''Complex Problem Solving'' (Do ¨rner 1980(Do ¨rner , 1996;;Frensch and Funke 1995a, b;Funke 1991;Funke and Frensch 2007).According to Buchner (1995, p. 28), laboratory studies on problem solving such as chess or the famous Tower of Hanoi ''… were criticized for being too simple, fully transparent, and static, whereas real-world economical, political, and technological problem situations were said to be complex, intransparent, and dynamic.'' To realize such complex situations under lab conditions, Brehmer and Do ¨rner (1993) recommended the use of computer-simulated microworlds.In the last 30 years, such microworlds have been used in numerous studies (for reviews, see Funke 1991Funke , 1995;;Funke and Frensch 2007;Osman 2009).There is a still ongoing debate if complex problem solving is a competence on its own (the related concept of operative intelligence is presented below), or if it is merely a subpart of ''intelligence'' (Do ¨rner and Kreuzig 1983;Kluwe et al. 1991;Kro ¨ner et al. 2005;Rigas and Brehmer 1999;Strohschneider 1991;Su ¨ß 1996Su ¨ß , 1999;;Su ¨ß et al. 1991;Wenke and Frensch 2003;Wittmann and Hattrup 2004).I will not answer this question here but point to one central question: Is complex problem solving a good case for complex cognition?Does dealing with complex environments require a special kind of complex cognition or is CPS possible by simply adding elementary cognitive processes?This question is important, because if there does exist a second type of cognition besides simple cognition, then many questions arise-for example, about the applicability of research results from the first area to the second or about the validity of theories from one area to the other.Before giving an answer, similar and related concepts to CPS like macrocognition and operative intelligence are presented and discussed shortly.

Concepts related to complex problem solving
The previous section showed that the term ''Complex Problem Solving'' is used in at least three different ways (thanks to an anonymous reviewer for clarification): (1) as a paradigm to study cognition under real-life conditions (with different foci such as learning, knowledge acquisition, and decision-making), (2) as a descriptor of behavior exhibited while dealing with a certain class of problems usually presented on a computer, and (3) as an ability construct that is related to intelligence.
To clear the different understandings of CPS, it seems necessary to elaborate its connection to other related concepts.I will consider in more detail the two concepts of (a) macrocognition and (b) operative intelligence.

Macrocognition
Pietro Cacciebue and Erik Hollnagel first introduced the term ''macrocognition'' in 1995.They used it as a descriptor for cognitive activities in natural decisionmaking settings.Klein et al. (2003) used it for contrasting a research program outside the lab with the traditional lab research on microcognition that is interested in the distinctive building blocks of cognition.
To illustrate their understanding of macrocognition, Klein et al. (2003, p. 82) mention phenomena like the following: planning and problem detection using leverage points to construct options, attention management, and uncertainty management.They contrast these issues with topics from microcognition research: puzzle solving, strategies for searching problem spaces, serial versus parallel processing models, or estimating probabilities of uncertainty values.
To be clear: the main difference between both types of research is not the time axis of the focused processes, but the ill-defined conditions of the tasks and problems under consideration.And to clear another potential misunderstanding: both approaches (''types of description'') are seen as complementary to each other.
With respect to the question ''inside or outside lab?'', proponents of the macrocognition approach seem to believe that these processes can only be analyzed outside ''in the wild''-to quote the famous book title from Hutchins (1995).But is that really true?I do agree that many lab studies use simplified experimental paradigms for testing hypotheses about specific processes.The reason for this procedure comes from the striving for internal validity (see, e.g., Shadish et al. 2002), which guarantees correct evaluation of postulated causal relationships.But, what about ''user labs'' or simulators (Gray 2002)?These setups allow for much more external validity in the lab, at the same time giving control over complex conditions and processes (Brehmer and Do ¨rner 1993).Additionally, progress in methodology allows for better data analyses in nonrandomized field situations (Shadish and Cook 2009).From my point of view, lab research is not necessarily restricted to microcognition but could be opened for macrocognition; also, macrocognition could be studied in field situations.Due to new analytic developments from our methodology, more accurate estimations of effect sizes can be used than before for analyzing such data.Also, by means of ''ecological momentary assessment'' new ways of data collection in real environments become possible (Hoppmann and Riediger 2009).

Operative intelligence
In an influential paper, Do ¨rner (1986) developed the concept of ''operative intelligence'' as a collection of those factors that determine strategic processes like flexibility, foresight, circumspection, or systematic behavior.His main point is a claim for the concentration on thinking processes instead of on their results.To a certain extent, the theory makes a micro examination of the details of information processing, while also considering a macro approach on action theoretical concepts.By using computer-simulated scenarios, as proposed by Do ¨rner and colleagues (Do ¨rner 1980), large datasets are collected which should not be directly reduced to a single score (''quantificatio praecox'').
Cogn Process (2010) 11:133-142 135 123 Author's personal copy Instead, changes and stabilities of the system variables as well as changes and stabilities in participants' behavior (asking questions, making interventions) should be identified.Thus, the term ''operative intelligence'' comes from an interest in the details of the psychic processes underlying CPS.
To give an example: Do ¨rner's (1986), p. 301ff) proposals for process indicators are divided in those related to questions asked by participants (only available if a thinking aloud method is used for data collection) and decisions made by them in the course of the simulation.Concerning questions, he differentiated between questions for states (''what is the value of … ?''), for dependencies (''what does … depend on?''), for effects (''what happens if … ?''), for components (''what do … exist of?''), for subordination (''what types of … do exist?''-specificformings of a variable), and for superordination (''what is … ?''-embedding of a specific object or event in a more abstract context).All these questions help a subject to identify the objects and their interrelations within the given domain.Concerning decisions, he distinguished the place of a decision from its dosage; for example, an intervention in the ''Tailorshop'' scenario could be made in the area of advertising, and the dosage could be the specific amount of money spent for that activity.Also, the sequence of decisions could contain interesting patterns, and the transitions from questions to decisions (and vice versa) are worthwhile analyzing.
The breadth of questions, the depth of exploration, the consistency of decisions, and the coordination of different action types are interesting indicators for describing the problem-solving behavior of subjects in complex situations.These indicators go far beyond variables that evaluate solution quality of a complex situation in a rather coarse manner.For example, analyses of transition frequencies between questions and decisions show significant differences between successful and unsuccessful problem solvers (Do ¨rner and Wearing 1995, p. 74f).
Both in traditional and in contemporary European research on problem solving (e.g., Buchner 1995; Do ¨rner and Wearing 1995), experiments are conducted using complex dynamic scenarios such as the commonly cited ''Lohhausen'' scenario (Do ¨rner et al. 1983).Sternberg (1995) states that this contrasts American research on problem solving which is mainly concerned with static paradigms (e.g., Chi et al. 1982) despite some exceptions (Anzai and Simon 1979).To come back to his theory of Operational Intelligence, Do ¨rner postulates four process components of dynamic situations, which were derived from empirical observations and are presented in the following section.
In order to successfully manage a complex system, (1) the agent has to gather information about the system and integrate this information into his model of the system.Typically, the system does not deliver any required information automatically.This forces the agent to ask specific questions or to conduct experiments that examine the way in which the variables work.This component of information retrieval and information integration is taken into account in the Multiple-Space Models by the experiment space.Goals in complex problems are often formulated only globally and unspecifically, for example, ''Manage this town successfully!''Therefore, the problem solver is required to render the nonspecific goal concretely.Additionally, goals might contradict each other, which makes it necessary to balance them by working out compromises or abandoning goals.Do ¨rner (1986) calls this (2) the ability of goal elaboration and goal balancing.To achieve this subgoal in turn, (3) the agent has to plan measures and make decisions.These measures might conflict with the agent's own system of values.This conflict can cause unpleasant emotions that affect the process of problem solving.Likewise, the agent has to cope with frustration, time pressure, and stress.In order to succeed in all this, (4) he or she needs to be capable of self-management.
The importance of Do ¨rner's contribution to European research on problem solving (e.g., Bach 2003; Do ¨rner 1986; Do ¨rner and Schaub 1994) cannot be overstated.Nevertheless, the theory of operational intelligence offers neither a fully elaborated theory nor a measuring device, although Do ¨rner repeatedly emphasizes the necessity of such a diagnostic device.Despite its stimulating character, the approach is not complete, resulting in a fascinating but loosely formulated theory, which still requires refinement.Because the status of a concept is sometimes best understood by pointing to the measurement approaches, the next section presents two assessment procedures from the field of CPS.

Assessment of complex problem solving
For the assessment of CPS, two different approaches exist (Buchner 1995): On the one hand, semantically rich ad-hoc simulations of complex microworlds (Brehmer and Do ¨rner 1993), on the other hand, formally constructed artificial systems, which allow systematic variation of difficulty (Funke 2001).As an example of the first approach, the system ''Tailorshop'' will be presented.As an example for the second serves the ''MicroDYN'' approach.
Tailorshop is a microworld, which realizes the production conditions of a small factory.The participant takes on the role of a manager of a small shirt factory.By means of changing the input variables (number of workers, number of machines, raw material, salaries, social security contributions, maintenance, advertising, shop location, utility van, and sales price), participants can increase the satisfaction of the workers, the production conditions, the demand for shirts, and the sales figures.Figure 1 shows the connection diagram of the 24 variables.
A monthly report gives the participant all the necessary information for making decisions whose effects show up the following month.-The system is build around a set of (non)linear equations dealing with variables representing the flow of money (e.g., into investments or from sales), the production (e.g., raw material, machinery, and workforce) and the turnover factors (e.g., prices, advertisement, demand, and sales).From a system analytic point of view, the scenario is centered on two subsystems, one being responsible for production (depending on machines, workers, salaries, social security contributions, and maintenance), the other one being responsible for sales (depending on price, demand for shirts, advertising, utility vans, shop location, and restricted by the number of produced shirts).
Pseudo-random processes (i.e., for all participants, the same sequence of random numbers) determine the price for raw material, the production, and the demand in previously defined value regions.For all variables, interventions into the system become functional in the next cycle, with one exception: the demand function becomes functional in the next but one cycle instead of a direct effect.

MicroDYN
A shortcoming of CPS research (as it was introduced by Do ¨rner in the 1970s) is its ''one-item-testing'' (Funke and Frensch 2007).Virtually all devices consist of one large and rather complicated scenario, the participant has to work through.At the end, either overall performance or various status and process indicators are calculated and evaluated.Thus, CPS instruments are tests which contain exactly one excessive item or at best one bundle speaking in IRT-terms (Embretson and Reise 2000) if various independent subsystems are considered as some authors do (i.e., Mu ¨ller 1993).Other tests allow subjects to explore a given system over a period of time and then ask several questions about this one system.That does not make the answers any less dependent.
Bearing these limitations in mind, the question arises how dynamic problem solving could possibly be measured with psychological tests.One possible approach to individual differences might come from the formal framework of linear structural equation systems (LSE-systems), which is called the MicroDYN approach (see Greiff and Funke 2009).This type of items has been used considerably in experimental research as indicators for problem solving performance (Funke 2001).The basic approach, however, is now a different one as outlined below.
Items based on this approach require participants to detect causal relations and control the presented systems.It is assumed that the everyday examples mentioned earlier can be modeled by MicroDYN systems, since advanced skills in strategic planning, internal model building, and system control are crucial in the specified situations as well Author's personal copy as those tested within the framework of systems.To solve the problem of one-item-testing, various completely independent systems are presented to the subjects (see below).In summary, Greiff and Funke choose to work within the formal framework of linear structural equation systems.The MicroDYN approach may be able to overcome some of the shortcomings mentioned earlier: (a) the lack of sound theoretical frameworks calls for a different kind of framework, which MicroDYN systems offer formally (theoretical foundation).(b) MicroDYN systems are easily constructed and can be freely varied in difficulty (scalability).(c) An infinite number of independent items can be presented (unlimited item generation).(d) Many everyday activities can be described by MicroDYN items (ecological validity).Because of these features and because of its excellent psychometric properties, MicroDYN has become the first choice for the PISA 2012 consortium for the world-wide assessment of problem solving.It helps to measure the ability to identify the unknown structure of artifacts in dynamic, technology-rich environments to reach certain goals.
MicroDYN systems consist of exogenous variables, which influence endogenous variables, where only the former can be actively manipulated.Possible effects include main effects, multiple effects, multiple dependencies, autoregressive processes of first order, and side effects, which all can be freely combined.Main effects describe causal relations between exactly one exogenous variable and exactly one endogenous variable.If an exogenous variable is involved in more than one main effect, this is labeled a multiple effect.Effects on an endogenous variable influenced by more than one exogenous variable are labeled multiple dependence.Participants can actively control these three effects as they manipulate the values of exogenous variables within a given range.Effects merely incorporated within endogenous variables are called side effects when endogenous variables influence each other, and autoregressive processes when endogenous variables influence themselves (i.e., growth and shrinkage curves).Participants cannot influence these two effects directly; however, they are detectable by adequate use of strategy.Additionally, all effects may differ in path strength.Figure 2 shows examples of the presented effects.
Participants see between 8 and 12 of these items each lasting about 6 min summing to an overall testing time of approximately 1 h including instruction and trial time.The MicroDYN items are minimally but sufficiently complex and at the same time sufficient in number.Each item is processed in three stages: (1) Stage 1, exploration phase: participants can freely explore the system.No restrictions or goals are presented at this time.Participants can reset the system or undo their last steps.A history to trace prior steps is provided.Exploration strategies can thus be assessed.(2) Stage 2, externalization of the mental model: simultaneous (or subsequent) to their exploration, participants are asked to draw the connections between variables according to their assumptions.This helps to assess acquired causal knowledge.(3) Stage 3, control phase: participants are asked to reach given target values on the endogenous variables by entering adequate values for the exogenous variables.During this phase, the practical application of the acquired knowledge is assessed.
To summarize: two measurement approaches, Tailorshop and MicroDYN, illustrate two different understandings of CPS.Whereas Tailorshop stands for a broad interpretation of CPS but has some weak points from a psychometric point of view, MicroDYN represents the psychometric sound realization of selected but important CPS aspects.In both cases, complex cognition seems necessary, because it is not merely a sequence of simple operations, which are however part of it, but a series of very different cognitive operations like action planning, strategic development, knowledge acquisition, and evaluation processes, all of which have to be coordinated and organized around the action goals.

The connection between cognition, emotion, and motivation
One of the fascinating aspects of complex cognition is its neatless connection to emotion and motivation.From our own experience with cognitive processes, we do not differentiate between cognition, emotion, and motivationinstead, we experience an integrated course of (internal as Fig. 2 Underlying structure of a MicroDYN item with all possible effects displayed.To the left are the exogenous variables, which can be manipulated directly by the participant.To the right are the endogenous variables, which have to be controlled by the subject well as external) actions with varying degrees will and varying degrees of affection.The driving forces behind any cognition and the affective states accompanying them are more than a by-product of cognition.The founder of the modern experimental psychology of motivation, see Lewin (1935), connected the dynamics of action with his concept of personality.Today, the neuropsychology of motivational processes is better understood (see, e.g., Kalis et al. 2008), but the integration of the separate modules is still an open question.Interestingly, in studies with simple cognitive processes emotions are normally blanked out, because they do not play an important role, for example in a visual discrimination task; but in studies with complex situations, the interactions become highly visible and inevitable: if the manager of the ''Tailorshop'' does not succeed with her investment strategy, bad feelings emerge which in turn influence information processing.In a recent experiment with ''Tailorshop'', Barth and Funke (2009) showed that ''nasty'' environments (where participants could hardly enhance the capital and received negative feedback most of the time) increased negative and decreased positive affect.The reverse was true for nice environments where it was easy to increase capital and to receive positive feedback.Furthermore, nasty environments influenced CPS by leading to a higher information retrieval and to a better CPS performance than nice environments.
At least three different approaches-the ''affect as information'' approach by Schwarz (1990), the ''assimilation-accommodation'' approach by Fiedler (2001), and the ''affect-infusion'' model by Forgas (2001)-compete with each other to explain the interaction between cognitive processes and affective states.All three approaches demonstrate the necessity to connect cognitive processes with affective states.In a more elaborated phase of theory development, instead of generally speaking about positive or negative affects, it might be even possible to qualify the differential effects of different emotions (e.g., happiness or proudness; fear or anger) on information processing and vice versa.Research on complex cognition would miss its goal if this interaction would not be an integral part of the next generation's theory.Especially, the aspect of emotion regulation (Koole 2009) might become a promising area.

Cognition: simple versus complex
Is there a clear border between simple and complex cognition, and if yes, how can one determine the degree of simplicity or complexity of a cognitive process?The founder of Experimental Psychology, Wilhelm Wundt (1832-1920), believed that psychological processes could be reduced to their basic elements (and provoked strong critique for this proposition by Gestalt Psychologists)-according to that position, complexity could be computed by simply counting the required elements.But what is the basic element of a psychological process?Is it a transition between two neuronal states?The overwhelming complexity of neural processes prohibits such a simple approach.
A primitive complexity indicator could be reaction time (Welford 1980): the longer the time needed for an operation the more complex must be the underlying process.This statement assumes the sequential processing of composed elements-a proposition which is questioned by many parallel processing models of cognition (see, e.g., Bechtel and Abrahamsen 2001;Mcclelland and Rumelhart 1985).To make things even more complicated, issues for psychological research have a broad range of time.Newell (1990) differentiated between 12 time scales of human actions, coming from four different ''worlds'' (see Table 1).
The Biological Band is currently more intensively focused than before due to the progress of neuroscience; the Cognitive Band has been an issue for microcognition since the Cognitive Revolution; and the Rational and Social Band could be the ones for macrocognition.
But, does this differentiation on the time scale really help?What else besides time could be used for discriminating simple from complex processes?Instead of concentrating on the cognitive process, a totally different approach would be to focus on the demands posed to the cognitive system by the task.Cognitive task analysis (CTA) has a long tradition in psychology (see Crandall et al. 2006).It tries to determine the thought processes that subjects follow to perform tasks at various levels of difficulty, from novice to expert.CTA looks at a specific system from the viewpoint of the subject performing a specific task.The number and variety of required tasks could be helpful in separating simple from complex cognition.Author's personal copy Another proposal, made by Halford et al. defines processing capacity by the relational complexity given by the number of related dimensions or sources of variation that can be processed in parallel.This approach would argue for a continuum between simple and complex processes.At the same time, it demands a definition of complexity (which is not the same as difficulty), which goes beyond the scope of this paper.

Concluding remarks
After a review of different constructs, some questions still remain open, e.g., what is the relationship between macrocognition, operative intelligence, naturalistic decisionmaking, and CPS?Is complex cognition the proper term or should not it be better labeled complex information processing?Concerning the relationship between the terms, it becomes obvious that they represent different research traditions; but at the same time, they all share the idea that there does exist a more complex interplay of cognitive processes than can be seen on the level of elementary processes.With respect to the question of labeling, the term ''complex cognition'' seems to be a good counterpart to the opposite term ''simple cognition'' even if it uses a broad understanding of cognition (including emotion and motivation) in the second case, which is not meant in the first one.As Moors (2009) argues, a broad sense of cognition contrasts the mental with somatic and motor responses, whereas the narrow sense contrasts cognition with motivation and emotion.The aim of this paper was to review the concept of CPS under the perspective of complex cognition.CPS differs according to not only my view markedly from simple problem solving by introducing additional facets of the problem space like intransparency, dynamics, or polytely.At the same time, there is also a difference between simple and complex cognition.Even if this is not a clear-cut concept with a sharp border, a major difference between simple and complex cognition lies in its interaction with different psychic functions like motivation and emotion which are strongly present in the complex case but missing or at least reduced in the simple one.This interaction requires additional regulation processes, which cannot be found when studying psychic functions in isolation and out of context.So in the end, from my point of view, complex problem solving turns out to be an interesting case for complex cognition.

Fig. 1
Fig. 1 Diagram of the connections within Tailorshop.Variables in bold boxes can be changed directly by participants (=exogenous variables), the others only indirectly (=endogenous variables)