Thinking about Thinking: The Relationship between Conﬁdence,

Attainment and Metacognition in Computer Science

Chris Napier

School of Computing, Newcastle University, Newcastle-upon-Tyne, U.K.

Keywords: Conﬁdence, Student Attainment, Metacognition, Undergraduate, Curricula Design, Modularisation.

Abstract:

Thinking about thinking, or metacognition as it is better known, is something that cannot be easily

taught but rather is developed and practised by an individual. It can be beneﬁcial because students make

connections between their learning experiences, discover their own learning preferences and can improve

their understanding of interconnection between concepts that the current HE system of modularisation tends to

undermine. It is a skill that an individual develops and practices over time based on a wide range of experiences

and learning opportunities. We hope to develop teaching approaches that help to improve students’ awareness

of their own metacognitive processes and learning strategies and by extension, their conﬁdence in applying

their previous experiences and knowledge to unfamiliar tasks and problems they encounter during their degree.

This paper outlines a preliminary study that explores the impact of conﬁdence on student attainment. The study

involved reviewing three stages of student experience - (i) The level of experience and conﬁdence of students

when entering the ﬁrst year of the degree, (ii) their attainment at the end of ﬁrst year and (iii) their level of

conﬁdence before and after coursework submission in the second year of their degree. Our results show that

there may be a direct relationship between conﬁdence levels and student attainment. Our results show there

is some link between metacognition and conﬁdence, with further exploration we can identify this link further

and create metacognitive learning strategies.

1 INTRODUCTION

On entering university, most students have a wealth

of knowledge and skills that are not purely academic

and have developed from their more general life

experiences. The learning from these experiences

can be used to improve or enhance their academic

performance but often students ﬁnd it difﬁcult to see

their relevance or potential to add value and so they

fail to transfer what they have learned to their

university studies. These skills can be relevant to

their learning in Higher Education (HE) but often

students ﬁnd it difﬁcult to see the connection and

transfer that learning to their HE studies. Flavell

(Flavell, 1976) ﬁrst introduced the term

metacognition and deﬁned it as being ”one’s

knowledge about one’s cognitive processes or

anything related to them”. Research in the area of

metacognition explores the idea that those who have

metacognitive abilities tend to be better learners and

have an awareness of their cognitive processes (Mani

and Mazumder, 2013) and thus tend to be better

learners.

As teachers, we want to ensure that each of our

students make the most of their learning

opportunities in HE and to perform to the best of

their ability. However, this is something that has

become challenging in recent years due to the growth

in student numbers. To ensure that each student gets

the opportunity to explore and develop equally we

need to develop tools and curricula that enable them

to take greater control of their own learning when we

are not around to support them, eventually leading to

them becoming the self-directed learners that we

would like them to be at the end of their degree

programme. This paper reports on a preliminary

study that explored to what extent stage 2 students

could see the theoretical and practical connections

between 4 different modules on their undergraduate

computer science programme and their conﬁdence in

completing the new module assignments based on

this awareness. Our initial ﬁndings conﬁrm that

some students could see the connections between

modules and that those students who could, had

greater conﬁdence and performed better in the

subsequent assignment.

212

Napier, C.

Thinking about Thinking: The Relationship between Conﬁdence, Attainment and Metacognition in Computer Science.

DOI: 10.5220/0007675002120217

In Proceedings of the 11th International Conference on Computer Supported Education (CSEDU 2019), pages 212-217

ISBN: 978-989-758-367-4

2 RESEARCH QUESTIONS

We used two research questions to explore the idea

that students who consciously use their

metacognitive abilities to transfer their learning from

one task to another are more effective learners and

are more conﬁdent in their ability to learn:

1. Can students see the relationship between

modules which develop similar skills between the

ﬁrst and second year of their degree?

2. Do those who are most conﬁdent in their ability to

learn perform better in their coursework?

3 BACKGROUND

3.1 The Effect of Conﬁdence on HE

Learning

There is a strong link in transitioning to HE and

conﬁdence. Lowe et al (Lowe and Cook, 2003)

investigated the question ”are students prepared for

higher education?” and found that 35% of the

students surveyed, felt they had chosen the wrong

course and 13% were unsure whether they should

actually be at university. This raises the question

whether the student had a lack of conﬁdence in their

own ability or whether there were external pressures

that made them choose their course. After two

months, the students were surveyed again and it was

found that students had a greater conﬁdence in their

choice and it increased to 57%. More interestingly,

the number of students who felt they shouldn’t be at

University increased to 19%. Lowe et al (Lowe and

Cook, 2003) do note in their investigation that a

substantial minority (20%) perceived themselves to

be lacking more in conﬁdence then they had

expected. This shows for the majority of students

that conﬁdence increased after 2 months of being at

University.

Previous experiences are strongly linked to

conﬁdence and there is extensive literature that

shows experiences within general CS concept and

programming greatly improve a student’s conﬁdence

(Alvarado et al., 2014; Hagan and Markham, 2000;

Bergin and Reilly, 2005) in university-level computer

science. Alvarado et al (Alvarado et al., 2014)

identify that most ﬁrst year HE courses are designed

to mitigate prior experiences. By designing a course

that mitigates prior learning experiences, all students

start off on a level playing ﬁeld in terms of

knowledge and the gap is closed in pre-requisite

knowledge needed to cope well in their degree. This

raises the questions whether the student can use their

cognitive processes to identify the relationship

between their previous experience and what they are

doing.

Denny et al (Denny et al., 2010) conducted an

experiment which asked students to predict their

performance in an introductory programming course.

Quizzes were held at the start of lectures to assess

their understanding of the material. Denny et al

(Denny et al., 2010) found that male students were

more conﬁdent than female students, which

coincides with current computer science education

literature. (Beyer et al., 2003; McDowell et al., 2006;

Sankar et al., 2015). Denny et al (Denny et al., 2010)

study also shows that female students achieved a

slightly higher (+2.6%) exam mark overall. Some

students are also overconﬁdent in their abilities and

this has been found in many studies (Gluga et al.,

2012; Pinto et al., 2017) which show that students

ﬁnd it hard to understand why they are

over-conﬁdent or even to detect that they are being

over-conﬁdent. In most of these studies the authors

found that over-conﬁdence was detected too late for

students to mitigate as they had already completed

their summative assessments. This raises the

question of whether previous experiences creates

overconﬁdence as students think they did well

previously so they are going to do well in the future.

By developing a student’s cognitive knowledge they

can understand the relevance of an

experience(Garner, 1987).

3.2 The Role of Metacognition within

Higher Education

Metacognition has two aspects, knowledge of

cognition and regulation of cognition (Schraw,

1998). Knowledge of cognition refers to the

knowledge that an individual has about their own

cognition and regulation of cognition is the set of

activities that can help an individual control their

learning. Garner (Garner, 1987) writes that those

who are said to be good learners are said to have a

better understanding of their cognitive knowledge.

The use of metacognition will be a way of helping

students to improve their learning. Metacognition

involves students having a deeper understanding of

their own cognitive processes (Flavell, 1976). By

helping students understand and improve their

cognitive processes, they will be able to have a better

understanding of their own learning. By developing a

deeper understanding of self-study and learning

transfer, students will be able to see the relation

between modules and should minimise the effects of

Thinking about Thinking: The Relationship between Conﬁdence, Attainment and Metacognition in Computer Science

213

modularisation.

4 METHODS

4.1 Participants and Setting

To explore the links between metacognition,

conﬁdence and student attainment, we reviewed the

marks and experience of 224 students in their second

year of study (three-year degree programme). We

looked at their exam and coursework results for the

ﬁrst year of study and their overall grade average and

also their initial results from one of their ﬁrst

modules in Year 2 - Operating Systems. After

reviewing the data only 221 students were included

in the ﬁnal review as some students did not sit the

Operating Systems exam due to personal

circumstances. We then monitored a sub-set of the

original population during the Operating Systems

module (29). These students volunteered to give

details on their level of conﬁdence before completing

coursework and afterwards, before receiving marks.

All participants in this study gave consent for their

results to be monitored and tracked over the course

of the programme. The ﬁrst year modules

(exam:coursework) used were Programming 1

(50:50), Programming 2 (60:40) and the Computer

Architecture (50:50) module.

The operating systems module has a weighting of

80% exam to 20% coursework. For this study we

chose to look at the coursework element as typically

students score very low in the coursework element.

The coursework element of the module has 2

assignments and we chose to speciﬁcally analyse the

ﬁrst assignment as it focuses on the core introductory

theories of the module. The ﬁrst assignment has 2

parts to the coursework in which part 2 can only be

completed by completing part 1.

4.2 Conﬁdence Level Measures

We gathered student conﬁdence levels through a

questionnaire. We used a modiﬁed Likert scale to get

the conﬁdence levels on individual questions. The

students could choose from the following: strongly

agree; agree; unsure and disagree. The questionnaire

was given out before the students attempted the

coursework and after the students had submitted the

coursework. The second questionnaire was

distributed only to those who had completed the ﬁrst

part of the experiment.

Table 1 shows the questions asked to students

when assessing their level of conﬁdence. The

Table 1: Questions asked to assess their conﬁdence.

(1) I feel conﬁdent that I understand what the speciﬁcation

is asking me to do

(2) I know what the material from the module I need to know

before attempting the coursework

(3) I will do well in this piece of/part of the coursework

(4) I believe this coursework/part will help me with preparing

for my exams

(5) This coursework will test my understanding of the material

questionnaire had 3 parts to it; the coursework as a

whole, part one of the coursework and part two of the

coursework. The questions in table 4 show what was

asked for the coursework as a whole and only slight

changes were made in the tense of the question.

5 FINDINGS AND DISCUSSION

5.1 CS1 Results against CS2 Module

When analysing ﬁrst year results, we wanted to see

the correlation between modules. We used Pearson’s

correlation co-efﬁcient to see the relation between

Programming 1 (CSC1021), Programming 2

(CSC1022), Computer Architecture (CSC1024) and

Operating Systems (CSC2025). The modules that we

had chosen to analyse had links in skills with one

another. In CSC1022, you needed to use the

knowledge and skills developed in CSC1021.

CSC1024 was analysed as CSC2025 requires

knowledge gained in the programming modules as

well as CSC1024.

Table 2: Correlation between Student results.

CSC1021 CSC1022 CSC1024 CSC2025

CSC1021 1 .738 .625 .426

CSC1022 .738 1 .656 .420

CSC1024 .625 .656 1 .568

CSC2025 .426 .420 .568 1

Table 2 shows the results from Pearson’s

correlation. Values closest to 1 show a strong relation

and values closest to 0 show no relation. When

analysing the results, we used the following guide by

Evans (Evans, 1995) to determine the strength of the

correlation: 0.00 - 0.19: very weak, 0.20 - 0.39:

weak, 0.40 - 0.59: moderate, 0.60 - 0.79: strong and

0.80 - 1.00: very strong.

When looking at the relation between CSC1021

and CSC1022 results (0.738), we can see there is a

strong relation in the results gained by students. In

order to do well in CSC1022, you need to understand

the concepts that are taught in CSC1021. When

looking at the correlation between CSC1022 and

CSEDU 2019 - 11th International Conference on Computer Supported Education

214

CSC2025, we can see that there is a moderate

relation (0.420). The programming language used in

CSC1022 is Java, which is a object-orientated

programming language. In CSC2025 the students are

not taught the programming language C, they are

expected to use their knowledge gained in CSC1022

to learn the language for themselves. This could

explain the moderate relation between the modules

as students cannot use their metacognitive skills to

adapt to a procedural language like C. When looking

at CSC1024, there is an increase in correlation

(0.568) but the strength of the relation is equal to

other modules.

Using Evans’ (Evans, 1995) guide we were able

to determine the strength of correlation between the

students results in CS1 and CS2. We found that there

was a strong correlation in results in the ﬁrst year

modules between one another but when comparing

the correlation to the second year operating systems

module, there was only a weak/moderate correlation.

This, at ﬁrst, looks like doubt is being cast on our

second research question but when you consider that

the majority of students improved on their result in

CS1 modules. CSC1022 is considered to be the most

important module on our programme, as it is the only

core module. Students who fail a core module

(¡40%) are not allowed to progress to CS2.

The improvement between CS1 and CS2 results

is a positive step in showing that metacognitive

processes are taking place.

5.2 Conﬁdence in CS2

We gave each part of the modiﬁed Likert scale a

value: Strongly agree: 2, Agree: 1, Unsure: 0,

Disagree: -1 and Strongly disagree: -2.

This meant that when looking at a student’s

conﬁdence, we were able to give them a value

associated with a speciﬁc part of the coursework and

the coursework as a whole, for both, before

attempting the coursework and after submitting the

coursework. We were also able to create a category

for the level of conﬁdence based on the value

returned: -10 to -5: Very Weak Conﬁdence, -4 to 0:

Weak Conﬁdence, 1 to 3: Some Conﬁdence, 4 to 7:

Strong Conﬁdence and 8 to 10: Very Strong

Conﬁdence. Using a categorical scale allowed us to

see if the conﬁdence of a student changed between

part 1 and 2. It is worth noting at this stage that no

student expressed strongly disagree to any question.

First we analysed the conﬁdence of students in

part one of the coursework. We found that the

majority of students (62.1%) can be identiﬁed to

have a strong conﬁdence and the next substantial

result was nearly a quarter of the students(24.1%)

who were conﬁdent. When looking at the categories

which would show a lack of conﬁdence (very weak

conﬁdence), there were no students who were

identiﬁed. Students with very strong conﬁdence were

a small minority (10.4%) along with a smaller

minority (3.4%) for those who had a weak

conﬁdence in their ability.

Table 3: Conﬁdence measures both before attempting and

submitting the coursework.

Before attempting After submitting

1 2 Whole Total 1 2 Whole Total

1 5 -1 3 7 8 7 9 24

2 5 1 1 7 4 3 4 11

3 5 -3 3 5 5 -3 3 5

4 8 6 8 22 9 10 9 28

5 4 4 3 11 4 4 3 11

6 5 4 3 12 5 4 3 12

7 3 -1 -1 1 6 5 6 17

8 10 9 10 29 9 10 10 29

9 1 1 3 5 1 1 3 5

10 5 3 4 12 6 4 3 13

11 2 3 3 8 8 -2 4 10

12 2 3 2 7 2 0 0 2

13 7 6 7 20 4 5 5 14

14 5 5 4 14 5 5 4 14

15 4 2 1 7 6 -1 1 6

16 4 1 1 6 10 5 6 21

17 0 -2 0 -2 0 -2 0 -2

18 4 4 2 10 4 4 2 10

19 4 0 2 6 4 0 2 6

20 5 3 4 12 4 -1 3 6

21 6 5 4 15 4 2 4 10

22 3 3 1 7 4 -2 1 3

23 6 1 4 11 1 1 2 4

24 3 1 -1 3 -1 -1 0 -2

25 6 6 7 19 2 4 4 10

26 2 2 2 6 2 0 2 4

27 4 0 4 8 4 0 4 8

28 5 3 7 15 3 2 2 7

29 9 9 7 25 6 4 3 13

In part 2 of the coursework, the results shifted to

show that conﬁdence had decreased between part one

and two. The most substantial category became

’some conﬁdence’ (44.9%), rather than ’strong

conﬁdence’ (27.6%). Cases in which students

showed ’weak conﬁdence’ rose (20.7%) and again no

students were categorised as having a lack of

conﬁdence. In all cases, students were less conﬁdent

or equally as conﬁdent in the second part of the

conﬁdent to the ﬁrst part.

Looking at the coursework as a whole the results

did not change drastically. Again, the majority of

students were categorised as having ’some

conﬁdence’ (48.3%) in the coursework and a ’strong

conﬁdence’ (34.5%). The number of students

categorised as showing a lack of conﬁdence, again,

Thinking about Thinking: The Relationship between Conﬁdence, Attainment and Metacognition in Computer Science

215

was zero.

When completing the questionnaire after

submitting the coursework, there was some minor

changes to the students level of conﬁdence. Part one

of the coursework found similar results to the

conﬁdence levels students expressed in before

attempting the coursework. The majority of students

, again, were categorised as having a ’strong

conﬁdence’ (55.2%). Interestingly, the number of

students categorised as having ’weak conﬁdence’

doubled at this stage (6.8%) and the number who had

’very strong conﬁdence’ (17.3%) rose. In part two of

the questionnaire, again, the results stayed similar to

that of before attempting the coursework. There was

no clear majority in this section and both ’weak

conﬁdence’ and ’strong conﬁdence’ had an equal

(37.9%) share of the results. Those who had ’very

strong conﬁdence’ (6.8%) remained the same.

Conﬁdence on the coursework as whole had minor

differences but the number of students who had

’weak conﬁdence’ (10.4%) and were ’conﬁdent’

(48.3%) remained the same. Those who were

’extremely conﬁdent’ (10.3%) decreased and those

who had ’strong conﬁdence’ (34.5%) increased.

5.3 Conﬁdence and Operating System

Result

The results show that students who achieved a high

mark also expressed very strong and strong

conﬁdence levels. Some potential outliers could be

identiﬁed within this study, student 29 expressed they

had strong conﬁdence in the coursework but overall

in the module they attained 35%. This could be an

indication of an outlier or a student who doesn’t have

cognitive awareness of their learning and have

misconceptions of how their previous experience

could relate to their current experience.

The student that attained 98% in the module was

categorised with very strong conﬁdence and this

should be expected in a student who achieved such a

high result. What is more worrying is the student

who failed the module but expressed strong

conﬁdence in their work. This level of conﬁdence

could just be for the coursework, rather than the

module as a whole. In hindsight, this is something

that we should have considered when analysing their

conﬁdence in coursework against their result in the

module overall.

6 CONCLUSION

The results of our study show that students who have

greater levels of metacognitive abilities and

experiences are able to transfer their learning better

and achieve a higher result. Further research and

investigation into how metacognition can be

facilitated within HE is needed. We hope to be able

to start designing curricula that aids the development

of student’s metacognitive processes. This curricula

will hope to negate the problems caused by

modularisation, and improve a student’s cognitive

development and learning transfer. Computer science

is a ﬁeld that rapidly changes and students leaving

HE need to be able to adapt to these rapid changes.

Through developing students metacognitive ability

and helping them see the connections from one

learning experience to another, we can better prepare

them for these rapid changes.

7 FUTURE WORK

There are several different routes for future work. In

this investigation we looked at conﬁdence levels that

students had in coursework. This could be extended

to look at exam questions and assess a student’s

conﬁdence before and after attempting a question.

This would be a minor extension of this study, but a

more major extension would include assessing the

conﬁdence of the student’s in modules other than

operating systems. If we were to extend this study to

programming and computer architecture then we

could monitor the student’s conﬁdence level as it

develops over modules. This investigation focused

on the role of conﬁdence with metacognition, future

work will look at other cognitive processes within

learning, such as learning styles and how they relate

with student attainment. Looking at the conﬁdence

and metacognitive abilities of pre-entry students to

courses and also recent graduates would be an

interesting step to this research. With the spin of

looking a pre-entry students we could identify any

metacognitive practice gained from previous studies.

Work with graduate students would allow us to see

the skills gained in their studies and the impact those

studies have had on their job.

REFERENCES

Alvarado, C., Lee, C. B., and Gillespie, G. (2014). New cs1

pedagogies and curriculum, the same success factors?

In Proceedings of the 45th ACM Technical Symposium

CSEDU 2019 - 11th International Conference on Computer Supported Education

216

on Computer Science Education, SIGCSE ’14, pages

379–384, New York, NY, USA. ACM.

Bergin, S. and Reilly, R. (2005). Programming: Factors

that inﬂuence success. In Proceedings of the 36th

SIGCSE Technical Symposium on Computer Science

Education, SIGCSE ’05, pages 411–415, New York,

NY, USA. ACM.

Beyer, S., Rynes, K., Perrault, J., Hay, K., and Haller,

S. (2003). Gender differences in computer science

students. SIGCSE Bull., 35(1):49–53.

Denny, P., Luxton-Reilly, A., Hamer, J., Dahlstrom, D. B.,

and Purchase, H. C. (2010). Self-predicted and actual

performance in an introductory programming course.

In Proceedings of the Fifteenth Annual Conference

on Innovation and Technology in Computer Science

Education, ITiCSE ’10, pages 118–122, New York,

NY, USA. ACM.

Evans, J. (1995). Straightforward Statistics for the

Behavioral Science. Psychology Series. Brooks/Cole.

Flavell, J. H. (1976). Metacognitive aspects of problem

solving.

Garner, R. (1987). Metacognition and reading

comprehension.

Gluga, R., Kay, J., Lister, R., Kleitman, S., and Lever,

T. (2012). Over-conﬁdence and confusion in using

bloom for programming fundamentals assessment. In

Proceedings of the 43rd ACM Technical Symposium

on Computer Science Education, SIGCSE ’12, pages

147–152, New York, NY, USA. ACM.

Hagan, D. and Markham, S. (2000). Does it help to

have some programming experience before beginning

a computing degree program? In Proceedings of

the 5th Annual SIGCSE/SIGCUE ITiCSEconference

on Innovation and Technology in Computer Science

Education, ITiCSE ’00, pages 25–28, New York, NY,

USA. ACM.

Lowe, H. and Cook, A. (2003). Mind the gap: Are students

prepared for higher education? 27:53–76.

Mani, M. and Mazumder, Q. (2013). Incorporating

metacognition into learning. In Proceeding of the

44th ACM Technical Symposium on Computer Science

Education, SIGCSE ’13, pages 53–58, New York, NY,

USA. ACM.

McDowell, C., Werner, L., Bullock, H. E., and Fernald, J.

(2006). Pair programming improves student retention,

conﬁdence, and program quality. Commun. ACM,

49(8):90–95.

Pinto, G., Rebouc¸as, M., and Castor, F. (2017). Inadequate

testing, time pressure, and (over) conﬁdence: A tale

of continuous integration users. In Proceedings of

the 10th International Workshop on Cooperative and

Human Aspects of Software Engineering, CHASE

’17, pages 74–77, Piscataway, NJ, USA. IEEE Press.

Sankar, P., Gilmartin, J., and Sobel, M. (2015). An

examination of belongingness and conﬁdence among

female computer science students. SIGCAS Comput.

Soc., 45(2):7–10.

Schraw, G. (1998). Promoting general metacognitive

awareness. Instructional Science, 26(1):113–125.

Thinking about Thinking: The Relationship between Conﬁdence, Attainment and Metacognition in Computer Science

217