Encouraging the Development of Computational Thinking Skills through Structured Dance Activities (Discussion Paper)

This paper investigates whether aspects of structured dance and choreography may improve cognitive skills related to computational thinking. It also explores the relationship between performance on computational thinking tasks and motivation. Twenty participants,aged between 7 and 9 years old, undertook the Beginners Computational Thinking Test (BCTt) as a pre-test, took part in three dance related workshops then completed the BCTt as post-test and responded to an adapted version of the Intrinsic Motivation Inventory (IMI). Student performance on the BCTt improved significantly from pre-test to post-test for both boys and girls and seems generalised across the computational categories measured by BCTt. The study did not show a correlation between performance and motivation in girls but suggests that there may be some correlation between improvement in score and reported enjoyment in boys.


INTRODUCTION
There is currently a skills shortage in the area of programming so there is a need to explore ways of increasing the number of people studying and working in the discipline.Women are currently under represented so initiatives to encourage more girls into the area may help.Computational thinking(CT) skills are linked to success in learning to program [14].There is some work [13] suggesting that it may be possible to teach computational thinking skills through dance.Dance is included in many contemporary K-12 curricula including the Scottish Curriculum for Excellence(CfE) [22] and, in the Scottish context, traditional dance is currently part of both the Physical Education and Expressive Arts curriculum.This is a structured form of dance which uses clearly defined sequences of standard formations.These are frequently described using a graphical notation.Extra-curricular dance classes are also popular, particularly among girls [21].The Scottish curriculum includes Computing Science as a compulsory component in the early years and primary years and includes the objective of "Understanding the world through computational thinking".The curriculum does not offer an explicit definition of Computational Thinking but follows the general approach of [5].Csizmadia et al. do not provide a single definition of computational thinking but are careful to provide example materials which can be used "unplugged" i.e. without the need for a computer as well as computer-based resources.CfE encourages cross curricular links and uses country dance examples when discussing the understanding of sequence in early years.This project investigates the potential for linking dance activities with the development of computational thinking skills.It focuses on the development of these skills in children aged 7 to 10 years old.This age range is selected because because kinaesthetic learning is important at this stage of development [24] and also because there is evidence that by the time pupils reach secondary school many have already decided that programming is 'not for them' [1].

LITERATURE REVIEW
Computational thinking is an approach to solving problems and designing systems which draws upon concepts fundamental to computing.Since Wing [28] popularised the term a number of researchers have expanded upon the idea.There is currently no single accepted definition of "computational thinking" but the systematic review by [12] lists abstraction, algorithmic thinking, problem solving and pattern recognition as the top four skills identified by researchers with decomposition also being important.[7], who review CT focusing on K-12 education, also emphasise abstraction, pattern generalisation, decomposition and problem solving while Roman-Gonzalez et al. [19] add correlations between computational thinking and spatial ability.
There is some debate as to how product oriented the process need be but a significant number of researchers and practitioners favour a definition which is not governed by context and which does not depend on programming or any other domain specific knowledge.These definitions concentrate on transferable skills which can be applied to solving problems across domains rather than focusing solely on features specific to computer programming.
Problem solving, abstraction, algorithmic thinking, pattern recognition and visual-spatial ability are also terms which occur in context of physical education and dance while the concepts of abstraction, decomposition and pattern recognition occur in literature pertaining to dance notation.
Although evidence to support the general statement that physical activity is positively associated with academic performance in school age children is inconclusive, there is evidence to support the idea that targeted interventions can have a positive effect on particular aspects of academic achievement [3].
The systematic review undertaken by [25] summarises current evidence for the effects of physical activity interventions on cognitive and academic performance in children from the perspective of health and sports focused research.It concludes that there is strong evidence for beneficial effects of physical activity on aspects of maths performance.Specifically improvements are identified in performance on tasks related to representations and algorithm, establishing relationships and identifying pattern regularities and descriptive and representative models and on task related to fluid reasoning, visual processing and short-term memory.There is significant overlap between these and the factors identified as important for computational thinking.
Work in the area of embodied cognition adds to empirical knowledge on interaction between physical movement and cognitive performance.Proponents such as [23] argue that cognition is not independent of the physical body and that the body and sensorimotor system influence cognition.[2] find improvement on abstract cognitive tasks after relatively unrelated physical activity and [27] identify work ascribing improvement in performance in mathematical and other scientific subjects to physical activity and to the teaching of Computing.
A number of studies from dance literature show the potential for dance to improve cognitive function e.g.[9,18] with some pointing towards dance having specific properties which enhance cognitive function in a way that is not found for other forms of physical activities.Although much of the literature focuses on creative dance, [15] finds that structured forms of dance appear to help students develop convergent problem solving strategies and to increase their performance on pattern recognition tasks.
Dance descriptions are frequently transcribed and recorded using diagrammatic notation.[13] have found improvement in computational thinking through creative dance choreography and virtual interactions.[26] argues that the act of reading from a dance score helps children recognize important movement ideas or ,in the terminology of computational thinking, aids abstraction.[10] argues that dance notation helps students to decompose tasks into constituent elements and to see patterns and relationships.
Thus physical dance activity along with dance transcription appear to have the potential to develop skills related to computational thinking.

PROBLEM STATEMENT
This paper aims to address the question "Do structured dance and choreographic activities improve performance on computational thinking tasks?".It also asks "Is there a relationship between participants' motivation and their performance on computational thinking tasks?" and "Is there a relationship between participants' perceived competence and their performance on computational thinking tasks?"

METHODOLOGY 4.1 Research Design
Ethics approval was granted from both the university ethics committee and Glasgow Education Services Research Group (ESRG), which oversees any research activity within the city's schools, before commencement of the project.Participants were recruited through school.The practical constraints of working within a school setting meant that it was not possible to assign participants to random groups and expose each group to different treatments.Instead the study had a within subject quasi-experimental design.It addressed the research question through the analysis of pre-and post-test scores on the BCTt [29] and used a modified form of the IMI [16] to assess levels of pupil motivation.Participants undertook the BCTt as a pre-test, participated in three dance workshops and then undertook the BCTt as a post-test and answered an adapted version of the IMI.

Instruments
A review of literature was undertaken in order to identify a valid and reliable measure of computational thinking.This identified BCTt as a tool which showed high reliability for the assessment of computational thinking in the intended age group and which had been used in pre-and post-tests study designs for "unplugged" computational thinking activities.BCTt measures six computational concepts: sequence, simple loops, nested loops, if-then, if-then-else and while.Review of the tests and discussion with primary school teachers determined that BCTt seemed to be of an appropriate standard to be used as a diagnostic assessment tool in the middle stages of Scottish primary education.It had been validated in a Spanish educational setting [29] so an initial pilot study was undertaken which confirmed that the test performed as expected in a Scottish school setting.
Motivation was assessed using an adapted version of the Intrinsic Motivation Inventory(IMI), a validated self-report measurement intended to assess subjective experience [16].The authors of the IMI recommend constructing a version of the test based on selected factors relevant to the question being addressed and using the items from those factors, randomly ordered and rephrased as necessary.A version of the IMI has previously been used in the context of primary school children learning computational thinking by [11].The version used in this study drew on this to select initial factors.It used a five point Likert scale with emoticons [4,8].Factor analysis was carried out as recommended by [16] and a final set of nine questions were selected from the interest/enjoyment and perceived competence sub-scales.

Participants
Volunteer participants aged between 7 and 9 years old were recruited from a Primary 4 class in a Scottish primary school.Twenty nine pupils were initially recruited but illness and other absence resulted in 20 children, 6 boys and 14 girls, completing both pre-and post-tests.Consent was gained from both parents and participants prior to the commencement of the study.Teachers viewed the class as consisting of generally highly motivated pupils.Activities took place weekly over a five week period and were carried out within the school day.The timing of the tests, five weeks apart, conformed to the procedure reported in [29].

Variables
This study observed the effect of participating in series of dance related workshops on participants' scores on the BCTt test.The independent variable is the dance activity and the dependent variable is the BCTt score.Motivation was explored through responses to the adapted IMI.

Teaching implementation
The activities in this study did not involve the use of computers in any way but consisted of • initial administration of BCTt as pre-test.The pre-test was carried out in a classroom setting.The researcher introduced the project and then introduced the BCTt test using the protocol advised by the authors.• Workshop 1: participants were taught two Scottish Country Dance steps, a straightforward dance which introduced the structure of the genre 1 and four basic dance formations.
Teaching followed a traditional dance class format with steps and formations being demonstrated before students "marked them out" (i.e.moved through the dance with a focus on the patterns of movement rather than on performance) slowly then attempted to replicate the dance.Dance diagram notation [17] was then introduced and used to teach a second dance.Pupils danced the second dance.• Workshop 2: started with a recap of dance notation.Pupils then worked in groups to interpret instructions for an unfamiliar dance from diagrams and dance it.Another dance, introducing additional dance formations, was taught using conventional methods and pupils worked in groups to create a diagram of that dance using stickers with diagrammatic notation symbols.• Workshop 3: recapped notation and formations then pupils worked in groups to create original dances based on the formations they had learned and the constraints of the genre.Dances were transcribed to dance notation using a combination of stickers and drawn symbols then demonstrated to the class.and pencils; information sheets with details of formations, timings and symbols; physical objects representing dancers to manipulate; dance template sheets and stickers with dance notation symbols.They also had freedom to move and mark out formations at will.The practical activities were all group based with pupils working in sets of eight to dance the initial dances and then groups of around four to interpret diagrams and create dances.Pupils were allowed to self allocate with some minor adjustments by the class teacher.The composition of groups varied from workshop to workshop.
The experienced class teacher and at least one other adult, either a classroom assistant or student teacher, were present in addition to the researcher to help lead activities each session.The additional adults all had some previous knowledge of SCD but had not previously encountered the dance diagram notation used.

RESULTS
The mean mark for the pre-test was 18.7 (SD=4.207)out of 25 and median mark was 20.5 out of 25.These results were compared to published data for children of the same age available from [6].Due to differences in education systems results from the Scottish sample were compared with published results for both Portuguese Grade 3 and Portuguese Grade 4. Kolmogorov-Smirnov was used to test for normality in samples of more than 50 while Shapiro-Wilk was used for smaller sample sizes.This showed that results for Portuguese Grade 3 ( n= 158,p< .001 ) Portuguese Grade 4(n=143, p< .001 ) and Scottish Primary 4 (n=24, p< .001)were not normally distributed.The Independent-Samples Median Test showed no significant difference in median value of Total Score for Portuguese Grade 3 and the Scottish Primary 4 group (n=182, p= .675)while the Independent-Samples Mann-Whitney U Test indicates that the distribution of Total Score was the same across the two categories (U=1929.50,p= .889).Similarly the Independent-Samples Median Test showed no significant difference in median value of Total Score for Portuguese Grade 4 and Scottish Primary 4 group(n=167 p=.076) while the Independent-Samples Mann-Whitney U Test indicates that the distribution of Total Score was the same across the two categories (U= 1326.00,p= .076).Scores for the three sets of students are shown in Figure 1.When pre-test results were split into the six computational concepts measured by BCTt the distribution of marks for pre-test followed a similar pattern to that seen in [6] for the same age group ( Figure 2 ).
In the post-test the mean and the median marks increased to 20.45 (SD=3.137)and 21.5 out of 25 respectively.Results were not normally distributed (Shapiro-Wilk p < .001and p= .016)so the non-parametric two-sided Related Samples Wilcoxon Signed Rank test was used examine the difference in scores.This found that there was a significant difference between the median marks for pre-test and post-test (p=0.007).
Median marks by computational concept for pupils who completed both pre-and post-test can be seen in Figure 3 and their mean marks are shown in Figure 4.A two tailed Wilcoxon Signed Ranks Test (Table1) shows no statistically significant differences in scores for individual computing categories.
The median mark in the post-test increased for both boys and girls (Figure 5).The Independent-Samples Median Test did not show a significant difference between gender in the medians of pupils'   The dance creation activity was group based with pupils allowed to self allocate.There was variation in individual performance within groups as shown in Figure 6.
This was primarily a formative exercise, encouraging students to apply knowledge about dance structure and notation.However dances were reviewed and evaluated based on the duration of the dance, the formations used and the progression of dancers.Figure 7 shows these grades alongside normalised median marks on the BCTt pre-and post-test for each group.Data was not normally distributed so possible correlations between pre-and post-test scores and marks for dance activity were investigated using Spearman.This showed that for groups with who took part in the pre-test, post-test and dance creation activity, there was a correlation between individuals' pre-and post test score (n=14 correlation coefficient   = 0.571 p =0.032) but no correlation between an individual's pre-test score and the mark awarded to their group for the dance activity (n=14   = 0.120, p =0.683) or an individual's post-test score and the mark awarded to their group for the dance activity (n=14,   = 0.250, p =0.389).When genders were considered separately there were no significant correlations between change in score and the four questions for girls but a positive correlation between change in score and Question 1 (n= 6,   = .853,p = .031)for boys.However, given the number of participants, caution is needed when considering this result.

Figure 8: Responses to IMI questions on interest and enjoyment
Perceived competence was measured using five positively worded questions as shown in Figure 9. Spearman's rho showed no significant correlation between the change in score and the answers to the questions on perceived competence (Table 3) .This remained true when genders were considered separately.

DISCUSSION
Student performance on the pre-test was in line with expectations for pupils of their age and educational stage.
Dance and transcription activities took place in self-allocated groups of students with differing levels of ability as measured by the BCTt test.The taught dance transcription exercises provided participants with opportunities to exercise skills in decomposition, abstraction, pattern recognition and visual-spatial recognition while the process of developing an original dance provided opportunities to exercise pattern recognition, visual-spatial and problem solving skills.
The dances chosen for inclusion were selected because they are composed of a limited set of dance formations which use recognisable notation symbols (Figure 10).Additionally the formations are relatively straightforward to code using visual programming tools, which may be pertinent in future studies.They are all categorised by the Royal Scottish Country Dance Society (RSCDS) as suitable for beginners.Pupils had some previous exposure to SCD and learned the selected dances successfully, with the dances providing a suitable level of challenge.
Interpreting the dance notation required high levels of concentration from participants.The related activities were group based and it was noticeable that when presented with diagrammatic representations of a dance most groups discussed possibilities while consulting formation sheets then marked out possible patterns rather than manipulating the physical objects provided.Accuracy of the transcribed dance diagrams produced varied.Two groups produced diagrams which were of the correct duration and logically sensible although containing minor syntactic errors.Another encoded three  out of the four phrases correctly.The remaining two groups encoded two of the four phrases reasonably logically but did not provide symbols for the other elements of the dance.These groups recognised that they had not provided a full transcription of the dance but ran out of time to complete the task.The dance creation activity involved each of five groups (A-E) creating a unique dance, transcribing it to paper and then demonstrating the dance to the rest of the class.Groups identified potential formations, discussed possible ways to link these together, moved sticker tiles around and marked out movements with reference to the formation sheets before committing the dance to paper.Again little use was made of the physical objects representing dancers.All groups produced a dance consisting of standard formations that were linked together successfully and which could be performed once through.All but Group C were of appropriate duration.Progression 2 provided a challenge but only one dance (Group E) could not easily be adapted to allow progression to take place.Group E transcribed their dance accurately.Group C had transcription errors in two phrases.Group B used notes to describe the last two phrases as the group ran out of time on the activity.The others had minor syntactic errors (see Figure11 for an example diagram.It was produced by Group A. Strictly speaking the final sticker should be rotated 90 degrees clockwise.However the meaning is unambiguous as the two symbols in the final phrase can only be interpreted in one way).Overall it can be said that, with one exception, groups created, recorded and danced a dance successfully.
Post-test performance on the BCTt improved significantly from pre-test for both boys and girls and appears generalised across all computational categories measured by the BCTt.There was no correlation between the change in an individual's test scores and the mark awarded to their group for the dance creation activity.
Participants who completed the Intrinsic Motivation Inventory interest/enjoyment questions were generally positive or neutral about the experience.The study did not show a correlation between change in score and motivation in girls but suggests that there may be some correlation between improvement in score and reported enjoyment in boys.This is consistent with findings in [13].The numbers involved in the current study were small so caution is needed when considering the results, but the relationship may be worth exploring further.It is possible that prior dance experience might be a factor.All participants had some prior experience of SCD in school.A number of the girls volunteered the information that they were currently participating in or had previously participated in extra curricular dance activities.No boys mentioned this.Further work to explore this is planned.
The study found no correlation between change in score and perceived competence.The study combined physical dance activities and transcription activities but questions were couched in general terms and did not seek to distinguish between activities.Future iterations will seek to distinguish between attitudes towards dance and attitudes towards transcription activities.

CONCLUSION
This was an exploratory study with a relatively small number of participants and care is needed when drawing conclusions as there are a number of possible confounding factors.When working with a pre-and post-test design there is always the possibility of a learning effect.To mitigate for this as far as possible the instrument selected for measurement was one which has been reported as reliable for use in pre-post studies [29] and the protocol laid out for its use was followed.There is also the possibility that the participants, who were all in the same school class, encountered other relevant materials during in the five weeks of the study.Discussion with the class teacher suggest that there were no specific materials which seem relevant but the possibility cannot be ruled out.Further work with pupils drawn from a more diverse population to avoid the possibility of all participants encountering relevant supplementary materials within the duration of the study is planned.
Workshops in this study were of limited duration and some activities were rather rushed.This may have impacted on some participants' learning.Duration and design of workshops will be revisited when future activities are being planned. 2 Dances are progressive with dancers starting in top place and either moving one place to second place, then third and then fourth place each time the dance is repeated or progressing directly from first to fourth place, then moving to third then second place.In all cases dancers start the dance in one position and finish in a different position each turn through.Progression is a concept that novice dancers, whether adults or children, frequently find difficult to grasp.
It should be acknowledged that the BCTt test does not claim to measure all aspects of computational thinking but focuses on computational concepts.It does not provide information about computational practices or computational perspectives.Investigations which include programming activity and broaden the evaluation to include performance on programming tasks are under development.The IMI questions used will also be refined to distinguish between attitudes to the different activities undertaken.
Overall the study suggests that the structured dance and transcription activities undertaken may have improved aspects of computational thinking, specifically computational concepts, and that further exploration is warranted.
• administration of BCTt as post-test and the IMI based questionnaire in a classroom setting.Each session lasted around an hour.Teaching was carried out by the researcher who is an experienced university lecturer, qualified school teacher and qualified SCD teacher.Music for the workshops was provided from recorded sources.Pupils generally demonstrated high levels of engagement.They were provided with a range of materials to support workshop activities.These consisted of paper1 Scottish country dances are typically 24,32 or 40 bars long and consist of a series of 8 bar phrases.Each phrase consists of one or more formations.Formations are 2, 4 or 8 bars long.Dances are danced in sets of four couples with dancers standing in lines facing their partner and are repeated eight times through.Dancers start the dance in one position and finish in a different position each turn.

Figure 9 :
Figure 9: Responses to IMI questions on perceived competence

Figure 11 :
Figure 11: Example of a dance created by participants

Table 2 :
Correlations between IMI questions on interest/enjoyment and change in BCTt score

Table 3 :
Correlations between IMI questions on perceived competence and change of BCTt score