Performance Indexes for Assessing a Learning Process to Support

Computational Thinking with Peer Review

Walter Nuninger

Univ. Lille, CNRS, Centrale Lille, UMR 9189 CRIStAL, F-59000 Lille, France

Keywords: Reflexive Learning, Formative Assessment, Problem Based Learning, Test Unit, Self-Evaluation,

Cross-Evaluation, Peer Review, Hybrid Learning, Digcomp.

Abstract: This article focuses on digital skills as defined in European Digital Competence Framework for Citizen

(DigComp). In the framework of a hybrid and blended course, a formative pedagogical scenario is proposed.

The training process consists of a formative situation of agile development of an application, supported by a

gradual process of evaluation with and by peers called SCPR. The proposal is the result of several years of

continuous improvement with engineering students enrolled in the IT module for non-developers. Learning

outcomes relate to Computational and Algorithmic Thinking (CAT). It is then possible to compare the impact

of our standard course design over several years with the group enrolled in full-time initial training between

2021 and 2023. A 3-index set, including the counter-performance index, enables us to analyse the effect of

the pedagogical device on learning profiles, and the evolution of positive feelings and difficulties experienced.

Qualitative data confirm the project's benefits and trainers' role in terms of student involvement and

perspective-taking, and provides information on the impact of the previous training path and the obstacles.

The proposed indicators confirm the pedagogical proposal and guide future prospects towards more relevant

indicators for monitoring CT learning within the DLE framework.

1 INTRODUCTION

The European action plan for 2030, “the way forward

for the digital decade” (EU no. 2022/2481), is aimed

at job retention, adult education and social inclusion.

This is pressing need due to the digitalization of

society, reinforced by the health crisis of 2019, which

has changed usage, but also the promising

developments embodied by artificial intelligence and

its integration into professions (including training).

Given the societal role played by future chartered

engineers in France, training programs must meet

these requirements. The learning outcomes cover the

full range of professional skills of the coach, leader

and manager, as well as the specific area of scientific

expertise. Firstly, collective and social intelligence

must be developed to enable beneficial interaction to

achieve a common, shared goal in disrupted

environments. Secondly, digital skills are essential

for using digital technology in professional practice

and learning to meet employment challenges. Since

1900s, issues of quality and learning performance in

https://orcid.org/0000-0002-2639-1359

higher education have contributed to the promotion of

active pedagogies and competency-based approaches

(Gervais, 2016).

The pedagogical scenario presented in this article

fits into this framework to support the computational

thinking (CT) of non-development engineers. This

hybrid blended-oriented course integrates a formative

situation (Raelin, 2008) to which a progressive

formative evaluation process is attached (Nuninger,

2024). The aim is to encourage involvement and

collective intelligence for team-based learning based

on a shared project. The learning-by-doing approach

calls for an active posture on the part of the students

and support (Grzega, 2005). The trainer explains the

concepts, guides without solving problems, but

corrects to ensure production conformity providing

feedback (Hattie & Timperley, 2007). Nicol &

Macfarlane's (2006) 7 principles enable tutors to

support self-directed and reflective learning: clarify

expectations, promote self-assessment, provide

quality feedback, encourage communication and

positive mindsets, and enable improvement.

Nuninger, W.

Performance Indexes for Assessing a Learning Process to Support Computational Thinking with Peer Review.

DOI: 10.5220/0013361100003932

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 17th International Conference on Computer Supported Education (CSEDU 2025) - Volume 2, pages 815-822

ISBN: 978-989-758-746-7; ISSN: 2184-5026

815

1.1 CT and Chosen Curriculum

According to Shutes et al. (2017), computational

thinking is “the conceptual basis needed to solve

problems effectively and efficiently (i.e.,

algorithmically, with or without the help of

computers) with solutions that can be reused in

different contexts. (It is) a way of thinking and acting,

which can be demonstrated through the use of

specific skills, which can then become the basis for

performance-based assessments of numerical skills”.

This reflects varied learning goals and priorities that

go beyond the knowledge object (Baron et al., 2014).

The deployment of artificial intelligence only

reinforces the need for data, safety and networking in

a world in transition. The underlying expectations are

the use of digital technology for learning activity,

work and compliant digital production. The Digital

Competence European framework for citizen

(DigComp 2.2) meets such requirements through 21

capabilities, divided into 5 domains (Vuorikari et al,

2022) given in Table 1. The targeted skill levels for

engineers range from professional to expert (grades

4-8), covering intermediate, advanced and specialist

levels. The curriculum covers hardware, networks,

software, and data representation, along with

functional analysis, algorithm description language,

and application development methods for clean code

and easy-to-maintain solutions (Martraire et al.,

2022), such as test-driven development (iterative unit

testing and refactoring). The emphasis is not on

coding languages, but on development processes, i.e.

the concepts of abstraction covering data and

performance, generalization for digital transfer and

decomposition through valid algorithms as sets of

efficiently assembled instructions.

Table 1: skill repository and chosen CAT curriculum.

DigComp 2.2 Learning by doing and by using

Information

data literac

file format, meaningful naming,

structure of list, mob-

rogramming

Communication

Collaboration

LMS, share storage space (cloud),

ile mindset and RAD

Digital content

creation

Pair- and Peer- programming

Refactorin

and clean codin

Safet

Unit testin

and code review

Problem

solving

Top-down functional analysis, test

driven development, pseudocode

1.2 Formative Digital Production

At the heart of our proposal to support the

computational and algorithmic thinking (CAT) is a

digital production (Figure 1). It is a formative

problem-based project that motivates Kolb's learning

cycle (Kolb & Kolb, 2005): concrete experience,

reflective observation, abstract conceptualization and

active experimentation in a collective environment.

The progressive assessment process SCPR (Self

and Cross Peer Review) attached to it is also

formative, motivating questioning, feedback,

decision-making and skills transfer (Nuninger, 2024;

Thomas et al., 2011). The challenge engages students

in the training, then the collective supports their

learning (Falchikov, 2005; Sadler, 2010) thanks to:

self-evaluation to give meaning and learning

autonomy for personal effectiveness; cross-feedback

for trust and action, then empowerment in a deeper

learning act; and peer review as an active observer

through shared, fact-based assessment.

The project's phasing invites students to discover

the coding environment on their own, then to program

in pairs, sharing the workstation display (one does the

input while the other controls, both self-assessing).

They are obliged to carry out unit tests to clarify the

need, and then to review the code of the other batches.

Finally, mob- and peer- programming involve

assembling batches in an agile spirit with a view to

reaching a common consensus (delivery) prior to the

confrontation with competing teams.

Figure 1: digital production training process with attached

SCPR (above) and tutor’s role (below).

1.3 Goals and Focus

In this paper, we review the enriched standard

pedagogical scenario that integrates a formative

digital production and progressive Self and Cross

Peer Review process (SCPR) in the context of its use

to support computational thinking. From 2021 to

2023, it is deployed in the full-time initial training

(FIT) for engineers who are not computer developers.

Section 3 then recalls the data collected for the three

groups studied. The focus is put on the counter-

performance index (cpi) and the two score variables

(SFpf

: positive feeling; SFed

: expressed difficulty)

introduced by Nuninger (2024). Following the results

in section 4, section 5 focuses on the effect of the

device on the learning profile. This is reinforced by

unit testing. The final section concludes the paper.

CSEDU 2025 - 17th International Conference on Computer Supported Education

816

2 PEDAGOGICAL DEVICE

The CT module learning outcomes for FIT students

meet the requirements of chartered engineers and

DigComp2.2 (intermediate to advanced levels). They

must be skilled digital users, “able to use problem-

solving methods to define specifications and

collaborate with experts to effectively lead digital or

data-related projects, while remaining aware of the

constraints and limitations associated with

integrating digital innovations into organization and

communication”. The underlying competencies are

personal effectiveness and social intelligence. The

chosen curriculum covers the concepts listed in the

second column of Table 1. The teaching approach is

based on the use of digital tools and the production of

digital solutions within the framework of hybrid and

blended-oriented courses (Figure 2). The 36-hour

teaching unit program is divided into two equal parts,

finishing with the one devoted to the collective rapid

application development project. Upstream, 15 hours

of online self-training begin on the Digital Support

for Guided Self-Study (Nuninger, 2017), structured

as a sequence of activities synchronized by

completion tests on Moodle. The aim is to involve

students, to develop their learning autonomy and

organization, and to facilitate their grasp of the

chosen coding environment Scilab prior to the

project. The formative project focuses on the

production of an application based on the initial code

supplied, and the function packages to be developed.

The data structure is imposed with the call sequences

for conventional processing (read, write, add, delete,

search...). In 2021, unit testing was introduced, and

test-driven development reinforced this requirement

in 2023. The project generates 3 outputs:

 a set of skills with a level of expertise built

with the team and confronted with the group;

 individual reviews on experience that

facilitates learners to take a step back;

 and project productions that value the work.

Figure 2: Standard hybrid blended-oriented course.

Validation of learning outcomes is based on the

average of project and final exam marks (out of 20).

Social intelligence is not directly assessed, but does

have an impact on the results of the project-based

practical work. A positive effect of the activity should

be reflected in the final individual mark. The aligned

final exam is based on the following evaluation

criteria sorted by increasing level of difficulty:

 mastery of standards (IDEF0, pseudo-code);

 assignment, read/write, iteration, alternative;

 understanding of lists and structures (pointers);

 able to develop a digital solution by assembly;

 able to debug by data control (test unit);

 basic proficiency in Scilab for clean code.

3 EXPERIMENTATION

We are interested in the three FIT groups that started

a 3-year chartered engineer training in biotechnology

and agri-food sector in 2021, 2022, and 2023. The

CAT module occurs in the second semester of the

academic year starting in September. The mean age is

around 20.4 years (20 in 2022), with age ranges from

19 to 24 years (19 to 23 in 2022). In recent years, the

number of students has fallen (-34% with respect to

2021) with the ratio of women to men dropping from

6.4 to 2.2, and a change in the distribution of previous

training paths (Table 2). Two senior teachers are

involved in the teaching unit each year (Table 3), with

the course leader (SL1) and a permanent trainer (SL3)

who has replaced the temporary substitute since 2022.

Table 2: Group characterization with training path.

Groups FIT2021 FIT2022 FIT2023

flow ; women % 50 ; 86% 45 ; 74% 33 ; 67%

. technolo

7 7 1

. Bach. of Sc. 18 9 6

reparatory path 17 15 12

Preparatory class 8 14 14

Table 3: Groups Project by Trainers (Student Numbers).

Senior lect. FIT2021 FIT2022 FIT2023

SL1 leade

(

)

(

)

(

)

SL2 substitute 2

(

)

- -

SL3

ermanent - 2 gr. (24) 1 gr. (11)

3.1 Data Collection and Processing

The data are collected primarily for educational

purposes, as explained and carried out in Nuninger

(2024). Pre- and post-module questionnaires provide

the qualitative data from which score variables are

Performance Indexes for Assessing a Learning Process to Support Computational Thinking with Peer Review

817

constructed to complement the quantitative data from

the assessments. The compulsory personal evaluation

questionnaire at the end of the project makes it

possible to identify clear positive and clear negative

feelings, and the absence of opinions. The data are

anonymized once the various sources have been

linked, but contextualized by group and trainers.

Incomplete data that cannot be reconstructed, data

relating to specific situations, or data in insufficient

numbers are rejected for this study. Redundant

variables are not pre-selected for the regression study.

Data processing is done using Excel, Scilab, and R.

Satisfaction survey respondent rates vary from 47%

to 58% (max 29) depending on group and year. This

is why we mainly compare the years 2021 and 2023,

focusing on full groups and SL1' groups (Table 3) to

limit biases related to professional practice and style,

and insufficient or missing data in 2022.

3.2 Score Variables

The score variables (Table 4) introduced by Nuninger

(2024) are the normalized min-max sums over the

interval [1,2] of the measures (yes/no) relative to 4

sets of descriptors at the beginning (B) and/or finish

(F) of the course. Difference variables (prefix Δ)

show changes in learner responses throughout the

course. The first pair (pp, ee) describes the learning

profile, while the second (pf, ed) reflects the final

feelings after completing the course. By construction,

only a finite number of values are possible (Table 5).

Table 4: Descriptors sets and score variables (letter S).

Refers to Descriptors

Pedagogical

preference

(SBpp

; SFpp

)

pp1: listen to the lesson

pp2: prepare and ask in class

p3: teamwor

Evaluation

experience

(SBee

; SFee

)

ee1: personal assessment

ee2: assessed colleague

ee3: confronted in a tea

Positive feeling

(SFpf

)

pf1: motivating to be evaluated

pf2: rewarding to evaluate others

f3: peer review is useful

Expressed difficulty

(SFe

)

ed1: difficult to evaluate oneself

ed2: difficult to evaluate others

Table 5: normalized score variable (subscript n).

Score SFe

SFpf

, SB/Fee

SB/F

Final value

(possible

configurations)

1 (1)

1.5 (2)

2 (1)

1 (1)

1.33 (3); 1.67 (3)

2 (1)

3.3 Counter-Performance Index (CPI)

Table 6 explains the meaning of the counter-

performance index (cpi) based on its range, defined

as the ratio of the normalized scores of final difficulty

expressed (SFed

) to positive feelings (SFpf

). An

expressed difficulty does not necessarily mean a non-

positive experience, as both can be experienced in the

same way (=1). cpi only compares the two scores,

without judging the cause. The cpi normality study is

based on 6 classes centred on class 3 (Table 7) with a

mean width of 0.25 ranging from 0.075 to 0.5 (nested

mean method), taking into account the permitted

values of the index and the [min, max] number of

respondents in the groups to be compared ([7; 29]).

Table 6: Meaning of cpi range and values in [0.5; 2].

Range cpi=SFe

/SFpf

Underlines the

> 1 2; 1.5; 1.2; 1.125 difficult

felt; >SF

= 1 SFdf

= SF

uilibrium

< 1 0.5; 0.6; 0.75; 0.9

ositive feelin

; >SFe

Table 7: Meaning of chosen cpi classes (width, values).

cpi class width values (configurations)

1: more positive 0.250 0.5 (1); 0.6 (1); 0.75 (2)

2: felt

ositive 0.175 0.9

(

)

3: balance 0.075 1

(

)

, i.e., no difference

4: felt difficult 0.175 0.125

(

)

5: more difficult 0.325 1.2 (1) ; 1.5 (2)

6: much more 0.5 2 (1)

4 RESULTS

4.1 Evolution of Assessed Levels

Figure 3: Evolution of project and final exam grades (/20).

Analysis of the scores reveals a difference in final

levels, influenced by group building and trainers’

style, but which is difficult to specify due to changes

over the period. The gap between final exam grade

narrows compared to 2019 and 2020 (Figure 3). This

suggests, firstly, an effect of the pedagogical

CSEDU 2025 - 17th International Conference on Computer Supported Education

818

proposal, which enables a greater transfer of expertise

within the groups, as noted by one student in 2023:

“we learnt at different speeds, but the exchanges

within the group enabled us to progress faster”.

Secondly, the impact of trainers' professional style, as

highlighted in 2023: “I appreciate that you tried to

motivate us. You listened to our difficulties and didn't

let us give up at the beginning” or “we were less

helped during the project than during the course”.

4.2 Positive Feedback and Barriers

Based on open-ended questions, we have identified

clearly positive and clearly negative opinions; others

are neutral (Table 8). In 2023, for all trainers and

groups, there are fewer clearly negative responses,

while the rate of clearly positive responses increases.

Table 9 presents the main obstacles grouped

according to 6 main dimensions identified:

1. Documentation (understanding of goals);

2. Environment (code syntax, lexicon);

3. Beginner (level, heterogeneous group);

4. Communication (group organization)

5. Allocated time (want more);

6. Commitment (late, starting a priori).

Table 8: clearly positive versus clearly negative feelings

expressed (over the respondents; others have no opinion).

FIT Full SL1 SL2/SL3

2021 12-36 %

(

)

8-38%

(

)

15-35%

(

)

2022 20-61 %

(

)

10-71%

(

)

28-52%

(

)

2023 7 -86 %

(

)

6-89%

(

)

10-80%

(

)

Table 9: Obstacles expressed by respondents (sorted).

Barriers 2021 2022 2023

No opinion 88% 38% 19%

1. Documentation 8% 17% 16%

2. Environment 0% 13% 22%

3. Beginner 4% 15% 16%

4. Communication 0% 8% 9%

5. Allocated time 0% 2% 13%

6. Commitment 0% 6% 6%

The first 2 obstacles are linked to the challenge,

which consists in expressing the customer's needs and

appropriating the initial codes supplied. The next 2

reflect the influence of the group (heterogenous) and

the previous training paths which also explain the

obstacles linked to the chosen coding environment

(Scilab compared to Python, which some students

were familiar with). The last two reflect the

difficulties encountered due to lack of time, but also

the desire to complete digital production successfully.

4.3 Qualitative Feedback

In 2021, personal assessments show a more positive

trend than in previous years, but remain focused on

individual behaviour (fear, lack of commitment,

objectivity) and the teacher's responsibility in the

final grade despite an activity recognized as

federating (Nuninger, 2024): “Despite initial

difficulties, I got back on track, completed the work

(and became) more efficient in testing to find errors

in others’ code.”; “we're proud of the work we've

done in the time available”; “we were creative, (but

had) difficulties in coding”; “It interested me (but) I

would have liked more step-by-step lessons”;

“enriching experience but a complex organization”.

In 2023, satisfaction levels are higher and show a

real sense of perspective on work, results and

individual and collective responsibility: “I had coding

experience but I adapted to Scilab”; “despite

difficulties, novices were motivated and committed,

developing programming logic with unit tests”; “we

did a good job. There are still mistakes but we've

improved”; “The project gave meaning to the

course”; “The project is rewarding, motivating and

requires us to communicate well”; “As a novice, I'm

proud of my progress”.

But some negative feedbacks remain such as: “I

didn't enjoy the experience too much, perhaps

because of a lack of interest in IT and a lack of

involvement. I didn't feel I had learned much”, “I

found it frustrating to rely on others' functions to

progress. Their errors and omissions caused delays

(much like in the professional world, you might say)”.

The school's satisfaction survey highlights these

aspects (Figure 4), showing increased confidence in

expertise identified in individual assessments despite

a decline in final results (Figure 3).

Figure 4: school surveys and individual reports analysis.

4.4 Counter-Performance Index

The beneficial aspect of the proposed pedagogical

device increases further between 2021 and 2023

(Figure 5) with a cumulative frequency rate rising

Performance Indexes for Assessing a Learning Process to Support Computational Thinking with Peer Review

819

from 41% to 50% for cpi inferior or equal to 1 (Table

10). The asymmetrical shape of the histograms shows

two distinct populations. For the SL1’s groups, this

was the case in 2021, but not in 2023. The distribution

is normal with a cumulative frequency rate dropping

from 50% to 36% (Table 11). This conclusion differs

for the groups followed by SL2 and SL3, but the

numbers are insufficient to draw any further

conclusions other than the impact of trainers.

Table 10: counter-performance index in 2021 and 2023.

Full FIT

rou

in 2021 2023

Res

ondents

(

rate

)

(

58%

)

(

54%

)

eneficial:1, 2 (<1) 24% 33%

alance:3 (=1) 17% 17%

difficult:4, 5, 6 (>1) 59% 50%

mean

(

std

)

1.32

(

0.47

)

1.16

(

0.33

)

Table 11: cpi for the SL1’s groups in 2021 and 2023.

SL1’s

rou

s 2021

(

SL1

)

2023

(

SL1

)

Respondents (rate) 12 (50%) 7 (50%)

eneficial:1, 2 (<1) 33% 27%

alance:3 (=1) 17% 9%

difficult:4, 5, 6

(

)

50% 64%

mean

(

std

)

1.83

(

0.41

)

1.22

(

0.38

)

Figure 5: cpi histograms and cumulative frequencies for full

groups in 2021 and 2023 (29 and 18 respondents).

4.5 Performance Index and Grade

The analysis of end-of-course learning profiles

(SFed

, SFpf

) against final grades is tricky due to the

varied possible response configurations (due to score

variables definition). By plotting the cpi in

descending order, we can identify 3 to 4 meaningful

areas in 2023 (Figure 6):

 cpi>1: a high expressed difficulty does not

prevent a positive experience but probably

favoured a higher final exam mark (area 1);

 cpi<1: a high positive feeling is favoured by a

lower expressed difficulty (area 2),

 then the decay of the two score variables lowers

the final score, while increasing them tends to

improve the final score (area 3).

For the groups supervised by SL1, in 2023 (middle

curve) the final mark decreases along with the 2 score

variables (area 2, cpi<1), and even more so with

increasing difficulty expressed (area 4). This was not

the case in 2021 (top curve): indexes are decorrelated

from the final grade.

Figure 6: cpi, SFed

, SFpf

and grade (/20) for full group in

2023 (bottom) and SL1’s groups in 2021 and 2023 (above).

4.6 Key Variables Influencing Indexes

Among the p pre-selected inputs using linear squared

correlation coefficient, our recursive identification

process retains the relevant ones to describe (cpi,

SFpf

, SFed

); i.e., minimum J

(sum of residuals

divided by (n-p); n being the number of respondents)

ranging between 0.16 and 0.38 in 2021 and 2023. In

2023, positive feeling (SFpf

) depends on the

student’s age but not in 2021, then of Δee1 (difference

in self-evaluation experience). The difference of the

cross-evaluation experience (Δee3) remains the

relevant input of the expressed difficulty (SFed

), in

addition to the difference of pedagogical preference

for lecture (Δpp1). In 2021, counter-performance

index depended above all on the evolution of

pedagogical preference with respect to lecture

CSEDU 2025 - 17th International Conference on Computer Supported Education

820

(Δpp1) and then to teamwork (Δpp3), but in 2023, it's

primarily the previous training path, then Δpp1.

4.7 Influences on Grades

The principal component analysis (PCA) shows that

project and final exam grades are not influenced by

gender, but might depend on students’ previous

training paths, and therefore on the groups assigned

by the school, on which the project groups depend. In

2023 (Figure 7), the examination grade depends

solely on the project grade (0.43 correlation). In

previous years, however, the grade also depended on

age and/or previous training paths, with correlations

exceeding 0.25. This was also true for project grade

in 2021, but not since 2022, indicating a real benefit

of the project in improving student expertise in CAT.

Figure 7: PCA focus on examination and project grades.

5 DISCUSSION

The 3-index set analysis indicates that the expressed

difficulty results from students' increased awareness

of learning outcome expectations and their effort

required to resolve the project, while the positive

feeling arises from the project's success and their

sense of personal evolution. Students involved in the

project have higher levels for both indicators and a

better chance of passing the final exam (grade higher

than 10/20). Conversely, a low level of expressed

difficulty may reflect a false sense of mastery. The

proposal positively impacts learning performance and

satisfaction (Nuninger, 2024; Schein, 2013).

In 2023, at the end of the course, the net decrease

(corrected for the increase) in students' preference for

the classic course (pp1) is -6% for a preference

expressed at 93% at the beginning (-14% in 2021, for

88%) in favor of active learning (pp2-3). In 2023,

72% of students recognize their experience of cross-

assessment and peer review, compared with 64% in

2021 for similar starting values (56% and 57%

respectively). Rates rise to 83% in 2023 and 75% if

self-evaluation is included (62% and 68% at the

beginning). The net evolution of self-assessment

experience is +17% in 2023 (+11% in 2021). Peer

interaction enhances pedagogical understanding and

raises awareness of evaluation (Topping, 2009).

Early teaching of unit testing is beneficial, as it

provides feedback (Scatalon et al., 2019). Students'

personal assessments show increased knowledge and

confidence in their computer skills. Debugging

expertise has improved, although the final level is

lower than expected. The reasons put forward are the

persistent decline in the level of students recruited,

the impact of the health crisis and the increased

difficulty of the final exam with the integration of unit

testing. Unit testing is challenging and increases

cognitive load, especially for beginners due to the

limited time available (Garousi et al., 2020).

Our pedagogical approach does not always

mitigate the effects on students' grades of their

previous training, the school-imposed groups, and the

chosen pairings. In 2021 and 2022, previous training

path is less dependent on age, but in 2023, the

correlation is strong (-0.42, while -0.18 and -0.01 the

previous years), confirming better-targeted

recruitment. Observation during the sessions reveals

the generational evolution. In 2023, students no

longer focus on the grade, but really express a wish to

understand, do and succeed in the challenge. It is the

combination of age (positive feelings) and previous

training (cpi) that contributes to acceptance of the

pedagogical approach adopted and commitment, with

the risk of disappointment (difficulty expressed). The

trainer-tutor plays an essential role to compensate for

the heterogeneity of learners' profiles, but is limited

by classroom constraints (Sadler, 2010). One student

points out: “the amount of help given to the groups

should be more evenly distributed to ensure fairness.

It's hard to get all the groups with different concerns

on the same track”.

In 2024, to understand how students approach

digital production and develop computational

thinking (learning profile) during the project, first an

online Kanban aims to collect the following metrics:

time spent on tasks with version tracking (cycle time),

time elapsed before task validation (execution time)

and throughput (performance and productivity).

Second, abstract syntax trees can help compare the

algorithms of imposed function versions within a

group and between different groups based on clean

code criteria. Third, we are currently prototyping an

automated data collection solution in the coding

environment Scilab to identify coding processes. A

first experiment took place in November with a group

of apprentices to assess their level of acceptance of

data collection and identify any difficulties in

integrating the extension into Scilab. Main constraints

lie in GDPR, data safety and storage, and GUI. Our

device's instrumentation will enable comparison with

other studies on computational thinking, even those

using different coding environments like Thonny (a

Performance Indexes for Assessing a Learning Process to Support Computational Thinking with Peer Review

821

Python IDE for beginners) with additional data-

collection plugins (Marvie-Nebut & Peter, 2023). The

final objective is to propose indicators for monitoring,

guiding, and evaluating remotely (DLE).

6 CONCLUSION

The proposed standard teaching scenario focuses on

skills through blended-oriented lessons and a

formative digital production to develop

computational thinking. The peer review process

reinforces reflective learning. Despite the complexity

of unit testing, the approach improves understanding

of algorithms and their design, debugging skills and a

willingness to validate solutions, helping future

engineers gain perspective. According to data

collected between 2021 and 2023, difficulty is

strongly influenced by students' previous training

path, in line with their age and social intelligence. The

cognitive load of beginners can only be mitigated by

more time devoted to them during the sessions and

the professional style of the trainer-tutors; a

parameter that has not been explored. The 3-index set

(counter-performance index, score variables of final

expressed difficulty, and positive feeling)

demonstrates the effect of the device on learning and

postures, and helps in learning profile analysis.

However, it is not sufficient to fully analyze the

learning processes of computational thinking.

To this end, larger student flows are required to

overcome the limitations of this work, but the

proposed training scenario is stable. The priority is to

instrument the Scilab coding environment, then to

identify students' coding processes in computational

thinking, and to determine learning profiles using

relevant contextualized indicators.

REFERENCES

Baron, G.-L., Drot-Delange, B., Grandbastien, M., & Tort,

F. (2014). Computer Science Education in French

Secondary Schools : Historical and Didactical

Perspectives. ACM Trans. Comput. Educ., 14(2), 11:1-

11:27. https://doi.org/10.1145/2602486

Falchikov, N. (2005). Improving Assessment through

Student Involvement: Practical Solutions for Aiding

Learning in Higher and Further Education.

Garousi, V., Rainer, A., Lauvås, P., & Arcuri, A. (2020).

Software-testing education : A systematic literature

mapping. Journal of Systems and Software, 165,

110570. https://doi.org/10.1016/j.jss.2020.110570

Gervais, J. (2016). The operational definition of

competency-based education. The Journal of

Competency-Based Education, 1(2), 98-106.

https://doi.org/10.1002/cbe2.1011

Grzega, J. (2005). Learning By Teaching: The Didactic

Model LdL in University Classes.

Hattie, J., & Timperley, H. (2007). The Power of Feedback.

Review of Educational Research, 77 (1), 81–112. doi:

10.3102/003465430298487

Kolb, A. Y., & Kolb, D. A. (2005). Learning Styles and

Learning Spaces: Enhancing Experiential Learning in

Higher Education. Academy of Management Learning

& Education, 4(2), 193-212.

Martraire, C., Thiéfaine, A, Bartaguiz, D., Hiegel, F. &

Fakih, H (2022) Software craft (French) Ed : Dunod,

289 p., ISBN : 978-2-10-082520-2

Marvie-Nebut, M., & Peter, Y. (2023). Apprentissage de la

programmation Python : une première analyse

exploratoire de l’usage des tests. In J. Broisin, et al.

(Éds.), Actes de l’atelier Apprendre la Pensée

Informatique de la Maternelle à l’Université. APIMU

2023 (p. 1-8). https://hal.science/hal-04144206

Nicol, D. J., & Macfarlene-Dick, D. (2006). Formative

assessment and self-regulated learning: A model and

seven principles of good feedback practice. Studies in

Higher Education, 31(2), 199–218. doi:10.1080/0307

5070600572090

Nuninger, W. (2017). Integrated Learning Environment for

blended oriented course: 3-year feedback on a skill-

oriented hybrid strategy. HCI Int, 9-14 July. In: Zaphiris

P., Ioannou A. (Ed.) LCT, 10295, Springer, Cham. doi:

10.1007/978-3-319-58509-3_13

Nuninger, W. (2024). Hybrid and Formative Self and Cross

Peer Review process to support Computational and

Algorithmic Thinking. In CSEDU 2024 (Vol. 2).

SCITEPRESS (Science and Technology Publication,

Lda). doi:10.5220/0012712800003693

Sadler, D. R. (2010) Beyond feedback: developing student

capability in complex appraisal, Assessment &

Evaluation in Higher Education, 35:5, 535-550, doi:

10.1080/02602930903541015

Scatalon, L. P., Carver, J. C., Garcia, R. E., & Barbosa, E. F.

(2019). Software Testing in Introductory Programming

Courses: A Systematic Mapping Study. 50th ACM

Technical Symp. on Computer Sc. Educ., 421-427.

https://doi.org/10.1145/3287324.3287384

Shute, V. J., Sun, C., & Asbell-Clarke, J. (2017).

Demystifying computational thinking. Educational

Research Review, 22, 142-158. https://doi.org/10.1016/

j.edurev.2017.09.003

Raelin, J. A. (2008). Work-based Learning. Bridging

Knowledge and Action in the Workplace. (New and

revised edition). San Francisco, CA: Jossey-Bass.

Thomas, G., Martin, D. & Pleasants, K. (2011). Using self-

and peer-assessment to enhance students’ future-

learning in higher education. Journal of University

Teaching & Learning Practice, 8(1), 1-17.

Topping, K. (2009). Peer Assessment. Theory into Practice,

48(1), 20–27. doi:10.1080/00405840802577569

Schein, E. H. (2013). Humble Inquiry: The Gentle Art of

Asking Instead of Telling". Berrett-Koehler Publishers

Vuorikari, R., Kluzer, S. & Punie, Y. (2022), DigComp 2.2:

EU 31006 EN, doi:10.2760/490274, JRC128415.

CSEDU 2025 - 17th International Conference on Computer Supported Education

822