The dynamics of the creative process.

The connotation of the word ‘dynamic’ supports the concept of the creative process not being linear [2] p. 295 explained the creative process as “a succession of thoughts and actions that leads to novel and adapted productions.” Research into the creative process has progressed rapidly since [11] four-stage model of creativity and several models have emerged over the past few years (for a detailed overview, see [2]) despite the absence of consensus on the definition of the stages. Cropley and Cropley [12] have mentioned seven stages—preparation, activation, generation, illumination, verification, communication, and validation. In Wallas’ four-stage model, Sadler-Smith [13] proposes a five-stage model, with an intimation phase added between incubation and illumination. Sawyer’s emergent model [14] examined six subprocesses (intuition, idea emergence, iteration, experimentation, and exploration). More recent work by [3] examined the concurrent sequence between the subprocess of generation, selection, exploration, evaluation, refinement, comparison, synthesis, and application. Other researchers have explored the stages of problem recognition, idea generation, idea evaluation, and solution validation in creative problem solving [15]. Furthermore, for an understanding of the creative act in its entirety, it is necessary to identify similarities and differences between the processes that result in creative outcomes and those that result in noncreative outcomes [2, 6]. Lubart presented four hypotheses as possible explanations for the differences. Under the first approach, creativity and non-creativity can be viewed as separate constructs that lead to creative and non-creative outcomes. The second hypothesis doesn’t differentiate these two processes and considers a creative and noncreative process continuum. Third, depending on the quality of knowledge used, the same process can lead to highly creative, moderately creative, or non-creative outcomes. Lastly, these processes may entail the same stages and may also involve the same amount of time spent at each stage, with the only variation being the quality of execution at each stage. Mumford and colleagues also outlined four hypotheses regarding the differences between creative and regular problem-solving processes. First, creative problem solving (CPS) involves ill-defined problems, as opposed to regular problem-solving. Second, CPS allows higher degrees of divergent and convergent thinking wherein routine tasks permit applying previously known procedures and information to solve the problem. Third, the difference lies in the multiple cycles of divergent and convergent thinking involved in CPS instead of regular problem-solving. Finally, the CPS process involves reorganizing and restructuring the information, while the routine, non-creative process just recalls the information based on the existing knowledge. These differences, though, require further investigation as there is no empirical evidence that can be used as a comparison between CPS and the simple problem solving (SPS) processes.

Emotions and the creative process.

Similarly, understanding the emotional processes that contribute to creativity becomes essential as they influence higher-level cognitive functions during the creative process, such as perception of the stimuli, judgment, decision making, and reasoning [16, 17]. Phenomenology research into creativity and emotions shows a wide range of emotions occurring during the creative process across many domains. Several creative individuals, e.g., artists, musicians, scientists, designers, describe the feelings of mixed emotions such as joy, happiness, and pain during the long processes of working and reworking on the creative problem to realize the idea of the product or outcome [18, 19]. Some long-standing views have examined the role of positive emotions in enhancing cognitive flexibility and creativity [8]. Others have explored the relationship between negative affects compared to neutral mood states in promoting creativity by helping individuals be more focused, critical, and determined in producing a creative outcome [20, 21]. The third paradigm produced reliable evidence for ambivalent emotions association with creativity [22], where highly activated positive (excited) and negative (angry) states were linked to high creative engagement and deactivated positive (relaxed) and negative (discouraged) states were linked to lower creative engagement [23].

Notably, a few studies have also investigated the dynamics of emotions for different creative stages. Peilloux and Botella [24] for example, found that the creative stages of immersion, thinking, research, inspiration, and insight were associated with positive emotions and the stages of judgment, experimentation, and planning were associated with negative emotions. Most recently, Kumar et al. [25] examined the dynamics of emotions using the eight design subprocesses, where positive affects dominated the conceptual phase of the design process whereas negative affects dominated the embodiment phase of the design process.

Existing measures of the creative process.

Protocol analysis. Protocol analysis has emerged as an important tool for behavior analysis during the dynamics of the creative process [3, 5, 26, 27]. Concurrent verbal protocols or the think-aloud method involves participants verbalizing their thought processes while performing the task in real-time using their short-term memory [28]. In the retrospective protocol method, participants retrieve the information of the already completed task that has been stored in their memory. Both these methods have their own advantages and limitations. Gero and Tang [29] have emphasized that both these methodologies achieve similar results with respect to examining the process-oriented aspects of the creative design process.

Markov’s chain analysis. Recent years have seen a rise in utilizing the Markov’s chain analysis to study the creative subprocess sequence. A Markov Chain is a stochastic model for predicting, estimating, or guessing the result of an event based on the preceding state and its action [30–32]. Pringle and Sowden [5] compared shifts of the associative mode of thinking with the analytical mode of thinking during the processing of emotional input. Moreover, Kan and Gero [33] examined design protocols from the standpoint of the sequential order of the Function Behavior-Structure (FBS) processes using the first-order Markov chain for various design processes. Nevertheless, the Markov analysis can be performed on any coding scheme. Ergodic Markov chains have the property of achieving a single stationary state distribution as time progresses. Generally, it is represented as a row vector π whose entries are probabilities that add to 1, and given the transition matrix P, it satisfies (1) where π is referred to as the equilibrium distribution of a chain. Computing a stationary state distribution makes it easy to statistically compare different distribution probabilities.

Existing measures of emotions during the creative process.

Most of the published research in measuring creativity and emotions together has relied primarily on diary studies and experience sampling methods [34, 35]. Even though these methods offer greater ecological validity, they are costly, time consuming, and do not account for in-the-moment emotional experiences of the creators. A further step in this direction should include the use of various technological advances that enable the investigation of the physiological and behavioral measures of the creative process and emotions [36]. A recent study employed eight body postures to decode an individual’s emotions during the creative design process using Kinect and a machine learning classifier [37]. Several articles have also asserted that facial expressions can be used to study emotions during the creative process [17, 38]. Maybe the concept of measuring facial Action Units (AUs) can be utilized to map the corresponding emotions that arise at different stages of the creative process.

The research gap.

Several articles have asserted that facial expressions can be leveraged to study emotions of the individuals during the creative process. To the best of our knowledge, no study has examined the dynamics of emotions during the CPS process using facial expression data.

Consequently, the two primary goals of this study are to examine the process related difference between the CPS and SPS processes and to capture in-the-moment emotional experiences of individuals using facial expressions during the CPS process.

Objective of this study.

More specifically, the objectives of this study are -

1. To compare the CPS and SPS processes using Protocol analysis and Markov chains.
2. To classify different stages of CPS process based on automatic facial AU data combinations.

Tasks.

Two different sets of tasks, CPS and SPS, each comprising of three different activities, were used in this study. The CPS task consisted of three different tasks (Fig 1), a drawing task [39], a writing task [40], and a structure making task [41]. The diversity of these creativity tasks allowed more room for creative engagement. They were chosen owing to the fact that working on a variety of creative tasks and switching between them enhances divergent and convergent thinking by reducing cognitive fixation [42]. The SPS task consisted of a drawing task, a writing task, and a structure-making task that were tailored to demand the same amount of effort as solving the creative tasks (Fig 2). The difficulty levels of the two sets of tasks were assessed beforehand by eight judges who rated the creative components (1 = not at all creative, 5 = very creative) and difficulty levels (1 = very easy, 5 = very difficult) of the tasks on a 5-point Likert scale. The two tasks did not differ in terms of their difficulty levels (t = −0.323, df = 14, p = 0.751). In contrast, the CPS tasks were rated significantly higher on creativity as compared to the SPS tasks (t = 15.82, df = 14, p < .001).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. CPS group tasks.

https://doi.org/10.1371/journal.pone.0269504.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. SPS group tasks.

https://doi.org/10.1371/journal.pone.0269504.g002

Capturing the facial expressions.

Participants’ facial expressions were captured using a custom-made head-worn camera. The aim was to ensure that the frontal faceview was always in view of the camera, regardless of the participant’s head movements during the task. An AKASO V50 pro sports camera was mounted on the headgear facing the participant at an approximate distance of 12.25 inches from their nose tip. The face was also illuminated by a strip of LED lights embedded in the headgear, as shown in Fig 3. The participants’ facial expressions were recorded while solving the tasks for 30 minutes at the rate of 25 frames per second in 720p.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Facial profiles of the participant with headgear set up.

The left figure presents the 3/4th view of the face, while the right figure presents the frontal profile. In both instances, the face has been appropriately illuminated using LEDs.

https://doi.org/10.1371/journal.pone.0269504.g003

A manipulation check was performed to ascertain that the headgear did not interfere with the task performance by administering the same to 18 randomly selected matched participants who were not part of the main study. Nine participants were required to wear the headgear, while the remaining nine did not and were asked to complete the alternative usage test [43] where they were asked to provide ‘five different usages of a brick’ within two minutes. The z-scores indicated no difference between those wearing the headgear and those who did not on fluency (z = −0.97, p = 0.39), flexibility (z = −0.225, p = 0.87) and originality scores (z = −0.94, p = 0.39).

Segmenting the protocols.

The video of facial expressions was synchronized with the timeline of the three dome camera recordings from the beginning to the end of the activity and the audio recordings with the video timelines. Adobe Premiere Pro CC was used to trim and synchronize videos and audios using the frame-matching features. Audio MP3 for each participant was mastered and boosted for better sound quality. This was followed by transcription of the verbal utterances on a word file. The transcribed verbal utterances were further divided into small units/segments/moves based on one main verb per parse [26] along with the timestamps in the datasheet for each participant. The start and end of a verbal utterance were timestamped based on the time interval of the recorded video that the participants were referring to. After parsing the verbal utterances, video recording observations that were not reported as verbs or actions during the retrospective verbal utterance process were separately segmented and were added to the same excel sheets with timestamps. They can be categorized as task-related actions (e.g., reading the task description, sketching, making the prototype, arranging the raw materials, etc.) and other actions (e.g., standing up, relaxing the body, etc.). The segmentation of actions based on video observation became more robust by capturing details that were obscured by verbal utterances alone.

Coding the segments.

Following the review of the segmented protocols, seven categories, namely preparation, ideation, illumination, evaluation, verification, production, block, or other were developed as coding scheme with the attributes of the different modes of thinking based on the literature review. Each segment was assigned to one category only. Table 1 summarizes the coding scheme based on which a total of 18,355 segments were coded for all the 64 participants.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Details of information category, example segments, and code explanation from participants’ protocols, all coded into seven stages.

https://doi.org/10.1371/journal.pone.0269504.t001

When dealing with a large number of categories, coding requires a high level of reliability. Accordingly, after the lead author completed this exercise, an independent coder who was blind to the experiment but was proficient in coding schemes and creativity research coded 1476 segments randomly chosen from the CPS and SPS groups. Cohen’s kappa was computed for these 1476 segments to determine intercoder reliability. We observed strong agreement in 82% of the segments (k = .816, p < .001) after correction for chance. In the event of disagreements between the coders, the coding categories were discussed between the two coders in order to categorize the protocols into different stages based upon different modes of thinking. Upon resolution of the discrepancy between the coders, the codes were adjusted to be consistent across all protocols for each participant.

Markov’s analysis to study the transition between different stages.

Based on the coding for the segments in the excel sheet, we calculated the probability of transition from one state to the next for each of the eight categories, providing us with an 8 by 8 matrix for each participant. Using these probability matrices, a stationary state distribution matrix with one row and eight columns was created using R studio for each participant to represent the differences in cognitive effort exerted at different stages of the task solving process. The analysis of Markov’s stationary state distribution in this study was based on the work by Kan and Gero [33].

Transition state diagrams.

The transition diagrams provide a visual representation of the transition of events with their probabilities. Based on the analysis of Jeong’s work [54], the transition state from one stage to another, for each participant, the sum of these frequencies was taken, and a relative frequency was computed for each transition. This provided us with a single transition matrix containing 64 probabilities of the transitions for the CPS and SPS groups, respectively.

In these transition diagrams, the creative stages are represented by the nodes that are linked to the other nodes using directional arrows. These directional arrows represent the relative frequency from one stage to another and the width of the arrow represents the strength of the transitional probability. Numbers in the transitional state diagrams represent the probability of one stage being followed by another stage.

Facial expression analysis.

Automatic facial action coding system. Facial Action Coding System [55] is a rigorous and psychometrically robust system that utilizes facial muscle movements to represent them as action units (AUs) to describe facial activity. Emotion-specific AUs or the combination of AUs can be found in EMFACS [Ekman & Friesen, 1983 [Unpublished]]. Several researchers across different domains have used automatic facial expression recognition software to characterize a specific set of emotions [56].

This study leveraged the OpenFace 2.0 [57] program to extract the facial action units from the synchronized videos of all the sessions. Based on comparing the performance of automatic facial expression recognition tools, OpenFace has been shown to be superior to the high-paid commercial softwares, e.g., Nodlus FaceReader, and Affectiva in detecting automatic facials. AUs with better accuracies [58]. Another advantage of OpenFace is that it is available in the open-source domain (https://github.com/TadasBaltrusaitis/OpenFace). This program is trained on different facial datasets and uses a linear kernel support vector machine to recognize individual AUs. The system recognizes seventeen specific AUs (1, 2, 4, 5, 6, 7, 9, 10, 12, 14, 15, 17, 20, 23, 25, 26, and 45). This study used only 12 AUs and their combinations that correspond to seven emotions. Details regarding the combination of AUs related to specific emotions used in this study can be found in [Ekman & Friesen, 1983 [Unpublished], 59, 60].

For each participant, OpenFace generated a total of 45000 frames based on the video duration of 30 minutes. These 45000 frames (rows) contained continuous AU data from 0 to 5 (intensity of AUs) for 12 facial AUs, along with the timestamp and confidence level. In most cases, AUs is regressed with high levels of confidence, usually 98%. To increase internal validity, frames with confidence levels less than 95% were removed from the analysis. Fig 5 demonstrates the OpenFace graphical user interface while simultaneously collecting the variables in a spreadsheet.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. User interface of the OpenFace 2.0 program.

Based on video and image outputs, this program detects and extracts participant facial features such as facial landmarks, head pose, eye gaze, and facial Aus.

https://doi.org/10.1371/journal.pone.0269504.g005

Support vector machine to classify the creative stages based on facial AU combinations.

The dataset generated by OpenFace required further processing before classification, as the facial AU datasets were highly imbalanced (typically a result of an unequal distribution of classes within a dataset) with some AUs with 0 values indicating no movement of specific muscles, while other AUs displayed continuous values (above 0), indicating movement intensity of the muscles at different time intervals.

As part of the feature set in this study, seven different combinations of AUs are used to represent seven different emotions to predict the target class, comprising of six stages. During the data analysis, the datasets of ‘illumination’ and ‘others’ stage were discarded as ‘illumination’ occurs within a fraction of a second [61] and is, therefore, inaccessible by retrospective protocol analysis and ‘others’ did not contain tasks-related information. Further, by eliminating these stages, the model’s performance improved by reducing the data imbalance issue. The final dataset for the CPS and SPS groups used for the SVM classification is presented in Table 2.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. CPS and SPS datasets of labelled in six stages (rows) and 12 AUs (columns) for all participants, used for the SVM classification.

https://doi.org/10.1371/journal.pone.0269504.t002

Before the model generation in SVM, we used a random undersampling approach to balance our facial AUs dataset using python libraries. We leveraged the imblearn.undersampling_sampling package in python programming language to undersample the data and generate a new dataset. The RandomUnderSampler function results in random selections of examples from the majority class. The parameter ‘replacement’ is set to false to prevent the repetition of the same example that requires deletion from the training dataset. This is followed using the fit_resample function to under sample the dataset and thereafter, train_test_split function has been leveraged to split into the training and testing dataset. 80% of the data were split into training sets and 20% as test sets.

We used a polynomial kernel function with degree 4 using Python to classify the stages, as other kernels did not provide high accuracy in predicting the stages based on different combination of AUs. Beyond this degree, the model became saturated. Moreover, our SVM model utilized the C parameter to avoid misclassification of the training data. Decision function shape is used as an ovo (OnevsOne) classifier to serve as a binary classifier for all pairs of classes, as well as the training utilizes the fit function to obtain a fitting of the model to the input training instances while testing utilizes the predict function to make predictions about the testing instances.

Preparation stage and associated emotions.

The stage preparation observed moderate accuracy scores for the ambivalent emotions, happiness (67.65% and 64.73%), and fear (68.95% and 66.52%) during the CPS and SPS process respectively. Positive emotions at the beginning of the task led some of the participants to approach the problem with high levels of motivation. For example, P01 (CPS) during task A reported—“this question was very interesting and I was just staring at them to understand the question” and P15 (CPS) reported,—“I was totally excited so immediately I started counting the shapes.” During task B, P21 (CPS) reported—“I like literature and writing and I can use my knowledge so this was really exciting,” and P03 (CPS) reported—“this question was interesting as well as tough.” During task C, P16 (CPS) reported—“this task was full of creativity and fun,” and P17 (CPS) reported—“I thought why not first solve the 3rd task as it looks very interesting”. For some of the participants, preparation was accompanied by a sense of apprehension after appraisal of the task about failing to complete the task. For example, during task A, P27 (CPS) reported—“I was worried because I was losing on a lot of time in searching for the pen to write”; P10 (CPS) reported—“I was a bit skeptical and stressed when I read this question.” Nevertheless, some participants also experienced negative emotions when attempting to complete the tasks on time. For example, during task A, P06 (CPS) reported—“but these shapes were complex and I got scared to finish it on time.” During task B, P13 (CPS) reported—“basically this was a political question and I realized that this task was going to take a lot of time.” During task C, P21 (CPS) reported—“I was also thinking about how much time I have left for this task. Participants in the SPS reported negative feelings in response to the focus and attention demanded by the tasks. For example, P15 (SPS) during task A reported—“this one demanded focus and attention and I had to be careful in converting the shapes”; P31 (SPS) during task B reported—“so I was also calculating the values in my head after looking at these graphs, that’s why it was taking time for me to see and then write the answers”; P17 (SPS) during the task C reported—“this kind of task is like want you to be very attentive because you will have to measure everything and make them.” Alternatively, the other perspective could be perhaps due to the nature of the tasks in the SPS group that were not exploratory and did not demand high divergent thinking abilities. For example, P06 (SPS) reported during task A—“there was nothing much to think about this task”; P31 (SPS) reported—“actually this was a complicated problem due to the shapes and you can’t come up with ideas here.”

Ideation stage and associated emotions.

The ideation stage also involved ambivalent emotions. These observations are in accordance with previous findings [8]. Moderate accuracy scores were observed for happiness (73.81% and 70.19%) for the CPS and SPS groups, respectively. For fear, accuracy scores were significantly higher (76.05%) for the SPS group than for the CPS group (56.90%). Perhaps the gap between ideation and implementation of those ideas during the SPS process seemed farfetched due to the nature of the tasks that resulted in such a difference. Some of the participants dedicated their cognitive efforts on making richer associations with positive emotions, thereby facilitating flexibility. For example, P27 (CPS) during task A reported—“so at first glance I thought that it was something like Picasso’s work and I was like very happy that I have direction now”; P02 (CPS) during task A reported—“it looked like a Dali painting somehow from which I got inspired.” During task B, P15 (CPS) reported—“I felt really happy that I came up with this idea of one nation one flag.” While during task C, P06 (CPS) reported—“I was thinking of making something similar to the Eiffel tower.” Perhaps, for some of the participants, not being able to transform pre-inventive structures into a concrete mental representation or idea caused fear during ideation. For example, during task B, P33 (CPS) reported—“I was scared if I will be able to complete this vague idea of UN ka intervention with Pakistan.” In the SPS group, negative affects encouraged participants to develop self-efficacy necessary to reach the right solution, resulting in positive affect. The participant initially struggled with converting a few shapes in Task A, but later came up with an idea for transforming the shapes. For example, P02 (SPS) stated—“I was not sure here but then I came across a rough idea of converting these arcs into semi triangle that made me feel lighter”.

Block stage and associated emotions.

Block is the only stage that observed very high accuracy scores for negative emotions—anger (96.04% and 97.26%), fear (94.67% and 97.26%), sadness (95.01% and 95.95%), and disgust (96.90% and 95.57%) respectively for the CPS and SPS groups. This makes sense as there are several factors that can be attributed to an impasse, for example, creative fixation, fear of failure, or lack of faith in oneself [63]. For example, during the task A, P1 (CPS) reported—“this was a sad moment because I couldn’t think of anything at this moment”; P2 (CPS) reported—“I was not able to come up with an idea right away which pissed me off”; P11 (CPS) reported—“It was very irritating that why I am not able to move forward”. During the task C, P19 (CPS) reported—“this paper was frustrating me because there was nothing I can do with this”. It is interesting to note, despite this, that some of the participants learned to accept these negative emotions, helping them to persist and complete their tasks, while others switched to an alternative activity.

Several participants in the SPS group were stuck on task A’s shapes, specifically the third element with two semi-circular shapes and the eleventh element with a star shape. For example, participant P01 (SPS) during task A reported—“at this point I was confused and irritated about converting this circle or not”; P02 (SPS) reported—“then at this figure the confusion happened and I was fearing that I could make this one wrong”; P30 (SPS) reported—“this was not triangle not circle both which was frustrating to decode”. In addition, some participants had difficulty rolling an 81-cm paper in Task C and encountered an impasse. For example, P04 (SPS) reported—“First I was not satisfied and I did not understand what I have to do to make it 81 cms”; P31 (SPS) reported—“even after trying so hard it was not 81 cms which was irritating”; P22 (SPS) reported—“I was like there was no option to make the length bigger which was very sad”. Participants in both the groups either abandoned the tasks immediately and switched to the other task or worked on the task for some time.

Evaluation stage and associated emotions.

The evaluation stage observed moderate accuracy scores for ambivalent emotions, happiness (72.25% and 74.09%), pride (64.63% and 68.08%), fear (60.74% and 67.83%) and stress (59.72% and 68.58%) respectively for the CPS and SPS groups. This state, which requires focused attention to determine whether the idea is weak, is an emotionally taxing process. For example, P19 (CPS) during task C reported—“but then this was not standing which gave me a lot of stress.” P05 (CPS), during task A reported—“but something was not very right here and it was very disappointing.” Similarly, in the SPS group, P01 (SPS) during task A reported—“in between I was checking whether I am doing it correctly as I was little worried”; P11 (SPS) reported—“I was having a lot of problem and stress in drawing this circle on this line”. For task C, most of the participants utilized their cognitive efforts on measuring the length of the rolled sheet. For example, P13 (SPS) reported—“mine was coming a bit long which was mind boggling”; P27 (SPS) reported—“when I realized this one to be 10 cms short I was like shit…”

It is interesting to note that some participants appraised their results positively which perhaps resulted in experiencing happiness. For example, P03 (CPS), during task A, reported—“I just looked at this thing and it came out to be nice.” P22 (CPS) during task C, reported—“here, when it stood na, I was very happy”; P06 (CPS) reported—“here I was happy that this structure could stand”. Our findings are consistent with previous research that showed designers experience high levels of arousal and positive emotions during the verification stage [64].

Refinement stage and associated emotions.

For the state of refinement, the CPS and SPS groups scored higher accuracy scores for ambivalent emotions—happiness (85.81% and 83.49%), pride (81.58% and 82.13%), and stress (77.19% and 81.19%) respectively. During this stage, additional layers of work, refinements, and finishing touches were added. Some participants were successful in resolving the issues and thus experienced happiness. For example, P01 (CPS) during task A reported—“I felt extremely happy when I made more flames to provide a completeness in this photograph.” Nonetheless, there were some participants who struggled to resolve the issues, which resulted in additional time and stress. For example, P14 (CPS) during task C reported—“even after spreading these legs, my idea did not work that tensed me.” P19 (SPS), during task A, reported—“I shaded the rest of the shapes to be free from tension”; P18 (SPS) reported—“I was having problem in redrawing this circle again and again.”

Production stage and associated emotions.

This stage observed moderate accuracy scores for ambivalent emotions—happiness (74.35% and 74.09%), fear (60.55% and 65.71%), pride (70.29% and 71.31%), and stress (60.71% and 71.83%) for both the CPS and SPS groups respectively. At this stage, all the tasks required the participants to produce tangible outcomes to implement an idea they had already conceptualized. Perhaps, those who were successful in transforming their ideas into successful outcomes felt happiness and pride. For example, during task A, P17 (CPS) reported—“once I had this face idea, I was successful in making it here.” During task C, P02 (CPS) reported—“I was enjoying making firm base out of this newspaper sheet.” During task A, participant P18 (SPS) reported—“I kept drawing like this because I was enjoying it.” During task C, P22 (SPS) reported—“I folded the sheet into exactly three parts and woaahhhh!!”. Perhaps, some of the participants may have realized that they could not produce a quality outcome and, as a result, might have experienced fear and stress. The realization of this ongoing process of production differs from the process of evaluation where the produced outcome or an idea is verified. It becomes difficult to capture the evaluation of the ongoing production process during the retrospective think-aloud protocol.