Resea ch Pape
Recommended ci a ion: Rachma , A., Wa e son, C., & Lundq is , K. (2025).
Compa ing The Impac o Gene a i e AI Cha bo s on S uden s' Re lec i e Thinking
in Lea ning P og amming. In Kangaslampi, R., Langie, G., Jä inen, H.-M., &
Nagy, B. (Eds.), SEFI 53 d Annual Con e ence. Eu opean Socie y o Enginee ing
Educa ion (SEFI), Tampe e, Finland. DOI: 10.5281/zenodo.17631485.
This Con e ence Pape is b ough o you o open access by he 53 d Annual Con e ence
o he Eu opean Socie y o Enginee ing Educa ion (SEFI) a Tampe e Uni e si y in
Tampe e, Finland. This wo k is licensed unde a C ea i e Commons
A ibu ion-NonComme cial-Sha e Alike 4.0 In e na ional License.
Compa ing The Impac o Gene a i e AI Cha bo s on S uden s'
Re lec i e Thinking in Lea ning P og amming
A Rachma a,
1
, C Wa e son b, K Lundq is c,
a Vic o ia Uni e si y o Welling on, Welling on, New Zealand, 0009-0004-5811-5369
b Vic o ia Uni e si y o Welling on, Welling on, New Zealand, 0000-0001-9471-6015
c Vic o ia Uni e si y o Welling on, Welling on, New Zealand, 0000-0002-5514-1519
Con e ence Key A eas: Digi al ools and AI in enginee ing educa ion
Keywo ds: Re lec i e Thinking, P og amming Cou ses, Gene a i e AI, Cha bo
ABSTRACT
This ull pape explo es how a speci ically uned cha bo can suppo s uden s'
e lec i e hinking. Resea ch in e lec i e hinking is ela i ely uncommon in compu e
science educa ion. In addi ion, implemen ing e lec i e hinking in eaching can be
challenging o academics due o he need o s uden pe sonalized suppo .
S uden s also o en iew e lec ion as an addi ional assignmen a he han a way o
imp o e hei lea ning. Despi e hese challenges, p og amming cou ses can bene i
om suppo ing s uden lea ning h ough speci ic e lec i e ac i i ies o os e
s uden s' e lec i e hinking. Suppo ing s uden s' e lec i e hinking in p og amming
can be achie ed by igge ing he p ocess h ough e lec i e dialogue. A gene a i e
AI-based cha bo can acili a e indi idual and con ex ual con e sa ions. This s udy
in es iga es he po en ial o uning la ge language models o igge e lec i e
dialogue as a lea ning ool. This s udy examined he in e ac ion be ween ele en
pa icipan s and he use o gene a i e AI lea ning ools in a con olled labo a o y
se ing, wi h one-on-one in e ac ions be ween he esea che , pa icipan , and
cha bo s. An expe imen al esea ch design was conduc ed, including a con ol g oup
(Cha GPT) and an expe imen g oup (ou p oposed cha bo ). A mixed-me hods
app oach was employed, combining bo h quan i a i e da a om p e- es and pos -
es ques ionnai es using Kembe 's Re lec i e Thinking Scale wi h quali a i e da a
om obse a ions and in e iews o assess pa icipan s' e lec i e hinking. The
esul s did no show any s a is ically signi ican di e ences in sco es on he Kembe 's
Re lec i e Thinking Scale. Howe e , hema ic analysis o he quali a i e da a
1
Aga ha Rachma
A Rachma
Aga ha[email p o ec ed]
.
e ealed ha pa icipan s in he expe imen al g oup demons a ed mo e e lec i e
hinking beha iou s.
1 INTRODUCTION
Re lec i e ac i i y is no a new opic in Enginee ing and Compu e Science
Educa ion. In lea ning p og amming, s uden s ha e o apply p oblem-sol ing skills
such as easoning, ques ioning, and e alua ion. E ec i e p oblem-sol ing equi es
e lec i e hinking. S uden s who neglec o employ e lec i e hinking skills o en ind
hemsel es unp epa ed, lacking in planning and sys ema ic app oach, and ailing o
conside al e na i e solu ions (A cı, 2022; Vi ian e al., 2013). Howe e , s uden s
s uggle o unde s and he pu pose o e lec ion and i s ele ance o he cou se o
s udy. Re lec ion is o en encou aged as a w i ing assignmen ha equi es w i ing
and linguis ic skills (Chan e al., 2021; Chng, 2018).
Academics also o en s uggle o acili a e e lec i e ac i i ies in hei cou ses.
Academics ha e o plan, p epa e, o ganise, implemen , and p o ide necessa y
eedback. These asks a e e en mo e challenging in la ge classes. Mo eo e , no all
lec u e s o ins uc o s possess he knowledge o he skills o acili a e e ec i e
e lec i e lea ning ac i i ies o o imp o e s uden s' e lec i e hinking (Chan e al.,
2021)
The eno mous numbe o a ailable and open sou ces o p e- ained la ge language
models (LLMs) suppo ed by so wa e de elope amewo ks o build cha bo s has
c ea ed an oppo uni y o pe sonalized gene a i e A i icial In elligence (Gen AI)
ools (HuggingFace, 2025; Langchain, 2025; Ollama, 2025). The e o e, i is possible
o de elop gene a i e AI-based lea ning ools speci ically designed o suppo
s uden s in engaging in e lec i e hinking. This s udy in oduces a speci ically uned
cha bo o s uden s and analyses he e ec s on hei e lec i e hinking.
2 BACKGROUND
2.1 Re lec i e Thinking
Dewey de eloped he concep o e lec i e hinking as "making meaning” o gain a
deepe unde s anding o connec ions o o he expe iences and ideas (Cla à, 2015;
Dewey, 1998). This concep aligns wi h cons uc i is lea ning heo y, whe e
s uden s build hei knowledge in connec ion o hei exis ing knowledge
(Be ssane e & de F ancisco, 2021; Coope s ein & Koce a -Weidinge , 2004).
Kembe 's ools o assessing e lec i e hinking ha e been ex ensi ely used in he
ield (Kembe e al., 2000; Leung & Kembe , 2003). Kembe ’s esea ch iden i ies ou
(4) le els o e lec i e hinking in lea ning (Kembe , 1999; Kembe e al., 2000;
Kembe , McKay, Sinclai , & Wong, 2008):
1. Habi ual Ac ion: Lea ne s do ou ine asks wi hou unde s anding he concep
o heo y. Expe s who ha e done simila hings many imes may be a his
le el. I 's like second na u e o expe s, while lea ne s migh jus ollow
ins uc ions wi hou hinking abou i .
2. Unde s anding: Lea ne s show some e idence o comp ehension o he
concep o opic. They a e beginning o g asp he basics and can apply hem
in a p ac ical sense.
3. Re lec ion: Lea ne s a e able o apply he concep o eal-wo ld si ua ions,
accompanied by pe sonal insigh , and s a o ques ion and analyse hei
unde s anding and lea ning p ocess. A his le el, lea ne s demons a e
me acogni ion, conside ing hei own hinking and cogni i e s a egies
(Sa gen , 2015).
4. C i ical Re lec ion: This is he highes le el o e lec i e hinking, whe e shi s
in pe spec i e on undamen al belie s ela ed o key concep s a e a e and
signi ican .
S uden s a e equi ed o possess e lec i e hinking o guide and make connec ions
be ween abs ac and dynamic concep s in lea ning p og amming (Medei os e al.,
2019). The Kembe Re lec i e Thinking Scale is u ilised as an analy ical ool o
examine he obse a ional and in e iew da a in his s udy.
2.2 Cha bo s in P og amming Cou ses
Since he ise o he popula i y o Gen AI cha bo s such as Cha GPT, Claude and
Gemma, lec u e s' and s uden s' pe spec i es on u ilising gene a i e AI cha bo s in
eaching and lea ning a e o en di ided in o hose who see po en ial bene i s and
hose who a e opposed o using gen AI cha bo in educa ion (Denny e al., 2023;
Essel e al., 2024; Lau & Guo, Aug 7, 2023; Rachma e al., 2025)
Ou p e ious in e en ion s udy we e conduc ed in an in oduc o y compu e
p og amming cou se in he School o Enginee ing and Compu e Science.
Employing a mixed-me hods esea ch o analyse he esul o he in e en ion o
using gene a i e AI cha bo as lea ning ool. The s udy con i med ha wi h p ope
in oduc ion and p omp guidance, lea ne s we e mo e inclined o u ilise cha bo s
wi hou nega i ely impac ing hei lea ning (Rachma e al., 2025). Addi ionally, i has
been shown ha cha bo s can assis s uden s who equi e suppo while lea ning
p og amming. Mo eo e , he e is e idence o e lec i e lea ning in hei lea ning
p ocess while assis ed by cha bo s (Rachma e al., 2025).
To u he analyse he in e ac ion be ween s uden and cha bo , as well as he e ec
o u ilising cha bo on s uden s’ e lec i e hinking p ocess du ing lea ning, we
conduc ed a quali a i e s udy in a mo e isola ed en i onmen o educe ex e nal
ac o s ha migh occu in classes oom se ings.
3 METHOD
3.1 De elopmen o a Cha bo o Re lec i e Thinking
Building upon exis ing li e a u e and ou p e ious s udy (Rachma e al., 2025; Whi e
e al., 2023), we de eloped a Re ie al-Augmen ed Gene a ion (RAG) cha bo wi h
he La ge Language Models (LLMs) amewo k ha se es as an adap i e acili a o .
By simula ing he in e ac ion be ween he cha bo and sugges ed p omp s de eloped
in ou p e ious s udy, lea ne s a e guided h ough i e a i e ques ioning and esponse
cycles.
Ou de eloped cha bo beha iou was modi ied o gene a e ques ions ha align wi h
he lea ning objec i es o he inqui y. This acili a ed a s uc u ed and inc emen al
explo a ion o he subjec ma e , mi o ing a e lec i e dialogue.
The p e- ained LLMs we e selec ed and explo ed using in-con ex lea ning (ICL).
ICL enables apid adap a ion o new asks using ew-sho examples, wi hou
equi ing upda es. The gemma2:2b model was selec ed due o i s supe io
pe o mance a e applying ICL, as i s small size makes i mo e accessible.
The chosen p e- ained model was hen in eg a ed in o he RAG wi h con e sa ional
memo y o main ain he con ex o he con e sa ion, addi ional knowledge o Swi
p og amming languages was added, and an in-con ex lea ning o al e he
beha iou o he cha bo .
The de eloped cha bo was hen es ed wi h h ee academics be o e he labo a o y
s udy. The h ee academics we e ep esen ed by: one expe in p og amming
languages, ano he pa icipan who had s uggled o lea n p og amming, and a hi d
indi idual who had ne e lea ned p og amming. Based on he es ing p ocess, he
cha bo unde wen signi ican imp o emen s, pa icula ly in i s abili y o gene a e
a ge ed ques ions and b eak down complex opics in o manageable lea ning s eps.
3.2 Resea ch Design
The s udy was conduc ed in a con olled labo a o y se ing wi h one-on-one
in e ac ion be ween he esea che , pa icipan , and a cha bo . The e we e wo
cha bo s used, and ou een po en ial pa icipan s we e andomly assigned o wo
g oups: he con ol g oup (CG) and he expe imen al g oup (EG). CG was alloca ed
o use he comme cially a ailable cha bo , Cha GPT, while EG was assigned o use
ou de eloped cha bo .
The s udy employed a con enience sampling s a egy, which is a non-p obabili y
sampling me hod. Con enience sampling en ails selec ing s udy pa icipan s who a e
eadily a ailable and willing o pa icipa e. Pa icipan s we e ec ui ed using lye s
dis ibu ed wi hin uni e si y campuses. The c i e ia included no p e ious mobile
de elopmen lea ning expe ience and ha ing less han a yea o p og amming
expe ience. Po en ial pa icipan s we e p o ided wi h a de ailed explana ion o he
s udy, indica ing hey we e allowed o op ou a any ime.
The s udy consis ed o h ee 1-hou sessions ha ook place o e 3 weeks. Th ee
sessions o lea ning we e designed o help pa icipan s de elop a simple mobile app.
In each session, pa icipan s we e gi en a ask o sol e based on he p o ided
lea ning objec i es:
• Session 1: To unde s and he Use In e ace (UI) amewo k and c ea e a
simple UI.
• Session 2: To unde s and Swi p og amming language (Da a ype and s uc )
and imp o e p e iously buil UI.
• Session 3: To na iga e be ween UI.
Pa icipan s we e also encou aged o jus ollow hei usual lea ning habi s and will
no be e iewed o es ed based on hei abili y. Pa icipan s we e also encou aged
o ask he cha bo .
3.3 Da a Collec ion
In o de o in es iga e pa cipan s’ lea ning expe iences ollowing he s udy, we
employed a combina ion o obse a ional me hod and semi-s uc u ed in e iew as
ins umen s o da a collec ion. The ocus o he in e iew was lea ning expe iences,
s uggles, and how hey u ilised he cha bo . Obse a ion checklis s and no es we e
used as ins umen s du ing he lea ning p ocess o eco d pa icipan s’ in e ac ions
wi h he cha bo , documen iden i ied p oblems and ill in addi ional in o ma ion
ela ed o he lea ning p ocess.
The in e iew and obse a ion da a we e e alua ed using Kembe ’s p e ious
esea ch on e lec i e hinking le els and lea ning app oaches (Kembe , 1999;
Kembe e al., 2008; Leung & Kembe , 2003).
4 FINDINGS AND ANALYSIS
Da a was ga he ed by obse ing pa icipan s du ing he h ee-week lea ning pe iod
and in e iew session. These no es we e inpu ed in o NVIVO 15 o s a he coding
p ocess. The ini ial coding session used an open coding me hod, and hen codes
we e ee alua ed and u he coded in he second ound o he coding p ocess.
F om he coding p ocess, hemes ha eme ged a e di ided in o ca ego ies o
lea ning p ocess and e lec i e hinking. The hemes ha eme ged om
obse a ional da a show a lea ning p ocess expe ienced by he pa icipan s, which
a e: Lea ning di icul ies and Lea ning s a egy. Table 3 e eals ha pa icipan s in
bo h he con ol g oup (CG) and expe imen al g oup (EG) exhibi di icul ies in
unde s anding he code and syn ax. The EG g oup demons a es s uggles and s alls
in iden i ying wha o ask and whe e o begin de ining he p oblem hey a e acing (2.
Does no know wha o ask).
Table 3. Pa icipan s Lea ning di icul ies and s a egies
Themes
Sub- hemes
Con ol
g oup (CG)
n=7
Expe imen
G oup (EG)
n=7
Lea ning
Di icul ies
1. T ying o unde s and code & syn ax
4
3.71
2. Does no know wha o ask
1.14
2.57
3. To p ocess new o mo e in o ma ion
1
1.42
4. Misconcep ion
1.57
0.71
5. To b eak down ask
0.42
0.28
6. To unde s and he p og amming concep
0.57
0.42
Lea ning
S a egies
1. Find mo e in o ma ion om ma e ial gi en
2.85
2.71
2. Ask ollow-up ques ion ( o ins uc o )
9.42
8.57
3. Ask he cha bo
3.28
6.14
4. Ask o he answe ( o ins uc o )
2.71
0.57
5. T ial and e o
1.86
2.14
6. Decomposi ion
2.85
0.86
7. Visualisa ion
0.14
1.14
The numbe s show an a e age o obse able ins ances pe pe son
Due o hei lea ning di icul ies, he pa icipan s a e mo e likely o eques guidance
om bo h ins uc o s and cha bo s. The CG pa icipan s we e obse ed o mo e likely
o seek o he answe a he han guidance, compa ed o he EG pa icipan s.
The hemes o how he cha bo is being u ilised a e: Ge uns uck, Assis wi h syn ax,
Explana ion & cla i ica ion, Hinde lea ning p ocess, and Imp o e he lea ning
s a egy.
Table 4. Cha bo ’s Impac on Lea ning
Pa e n
Con ol g oup
(CG) n=7
Expe imen G oup
(EG) n=7
Ge uns uck
0.42
3.50
Assis wi h syn ax
0.71
2.16
Explana ion & cla i ica ion
0.28
2.71
Hinde ing lea ning p ocess
0.85
0.28
Imp o e he lea ning s a egy
0
0.33
The numbe s show an a e age o obse able ins ances pe pe son
Table 4 shows he impac o he cha bo on pa icipan s’ lea ning wi hin h ee
sessions. Pa icipan s wi hin he EG we e using he cha bo o Ge Uns uck, which
ep esen s he unc ions o cha bo s in p o iding p omp s o s a coding, explo ing
keywo ds ha can be u he examined and unde s ood, and o e ing clues a he
han immedia e answe s. Meanwhile, CG pa icipan s had ewe obse able
ins ances o asking o assis ance o Ge Uns uck. Mo eo e , mo e obse able
ins ances o he Hinde ing lea ning p ocess we e shown in he CG g oup. Examples
o Hinde lea ning p ocess: he cha bo p o ides a long esponse ha causes
pa icipan s o eel o e whelmed and unable o p ocess he in o ma ion and he
cha bo p o ides immedia e answe o sol e he ask gi en.
Table 5. Pa icipan s’ Re lec i e Thinking Le els Based on Obse a ion
Le el
Con ol g oup (CG)
n=7
Expe imen G oup (EG)
n=7
Habi ual Ac ion
5.4
3.14
Unde s and
2.14
3
Re lec ion
0.28
4.4
C i ical Re lec ion
0
0.27
The numbe s show an a e age o obse able ins ances pe pe son
Table 5 shows pa icipan s in CG ha e mo e obse able ins ances ha align wi h
lowe le els o Kembe 's Re lec i e Thinking Scale, and pa icipan s in EG ha e
mo e obse able ins ances wi h highe le els o Kembe 's Re lec i e Thinking Scale.
5 DISCUSSION
Based on he pa icipan s' lea ning s a egies, i was ound ha hey we e mo e likely
o seek guidance om ins uc o s because hey we e less com o able using
cha bo s o lea ning, which aligns wi h p e ious s udies ha men ioned s uden s
ha e conce ned owa ds gen AI cha bo o lea ning (de Ke eki & Ga ido, May 8,
2024; Rachma e al., 2025).
Fu he analysis o he in e ac ion be ween pa icipan s and hei assigned cha bo
e eals ha he EG u ilised he cha bo o o e come obs acles. Speci ically, i was
obse ed ha EG pa icipan s we e s uck on lea ning di icul ies and did no know
how o p oceed, bu by asking ques ions and u ilising he esponse p o ided by he
cha bo , hey we e able o con inue he lea ning p ocess. Fo example:
Pa icipan 3 asked “How o esize jpg image in swi UI?“ and pa icipan 9 asked
“How o pu a ec angle image?“, while pa icipan 11 copy pas ed he code o he
cha bo and ask wha is he e o . Pa icipan s 3, 9 and 11 men ioned he ques ions
gene a ed wi h explana ion and code examples p o ided by he cha bo p omp ed
hem o s a w i ing hei code and igge mo e ques ions o con inue hei lea ning
p ocess.
Ano he obse able in e ac ions in he EG g oup is o ge mo e explana ion o
con i ma ion. Pa icipan s 1 (EG) and 3 (EG) asked “Wha is swi UI?”, Pa icipan 1
u he inqui y abou modi ie in swi UI. Pa icipan 9 (EG) asked “wha is s uc in
simple concep ?” hen ied o connec he new in o ma ion wi h p e ious knowledge
by asking “is i he same concep o class in OOP?”. Mo eo e , Pa icipan 5 (EG)
men ioned “i explain wha he unc ion o did, howand wha he syn ax was ha i
ge s clea examples on he syn axes”.
The CG's pa icipan s exhibi ed ewe obse ed ins ances equi ing assis ance om
he assigned cha bo . This migh sugges ha Cha GPT was be e and i s abili y o
p o ide solu ions is a no able ea u e, howe e , ha caused he lea ning p ocess o
s op whe eas ou app oach was designed o encou age u he lea ning and
disco e y which led o a deepe le el o lea ning. This was exhibi ed by he ew
obse able ins ances om he CG ha demons a ed explo a ion and eques s o
Explana ion & Cla i ica ion om he cha bo highligh ing a shi in he lea ning
p ocess's ocus owa ds only p oblem-sol ing o e knowledge acquisi ion.
The pa icipan s in he EG we e obse ed o be able o manage hei lea ning
di icul ies and ha e mo e a emp in ying o unde s and which aligned wi h Boyd’s
concep o e lec i e lea ning ha men ioned e lec i e lea ning is he abili y o
na iga e s uggles(Boyd & Fales, 1983).
Fu he analysis o pa icipan s in e ac ions wi h he cha bo and ins uc o e ealed
mo e obse able ins ances o pa icipan s in EG engaging in highe le els o
Re lec i e Thinking beha iou : Re lec ion and C i ical Re lec ion le els in Kembe ’s
Re lec i e Thinking Scale. In con as , he CG pa icipan s emained a he lowe
le els o Re lec i e Thinking, speci ically Habi ual Ac ion and Unde s anding le el.
Kembe desc ibes Habi ual Ac ion as making no a emp o each unde s anding
while lea ning (Leung & Kembe , 2003), which is he lowes le el o Re lec i e
Thinking, commonly ound among no ice lea ne s. This beha iou aligns wi h
obse ed ins ances whe e lea ne s ollow ins uc ions wi hou a ull unde s anding o
he p og amming concep and less likely o seek mo e explana ion o a emp o
unde s and. Fo example, Pa icipan 10(CG) and Pa icipan 12(CG) we e no ed o
be asking ques ions solely o comple e he assigned ask, a he han a emp ing o
gain a deepe unde s anding. Pa icipan 9(EG) was obse ed copying and pas ing
he code p o ided in he ins uc ions and he code p o ided by ou cha bo . Simila ly,
Pa icipan 8(CG) ecei ed an immedia e answe o sol e he gi en ask as a
esponse om he Cha GPT.
13 ou o 14 pa icipan s showed he beha iou o he Unde s anding le el, whe e
pa icipan s a emp o unde s and a concep o opic. Kembe men ioned ha a his
s age, he concep s ha lea ne s a e ying o unde s and a e s ill abs ac and lack
pe sonal ele ance, making i di icul o apply in eal-wo ld con ex s (Kembe e al.,
2000; Leung & Kembe , 2003). Bo h he CG and he EG exhibi his beha iou as he
lea ning p og ess con inues. The obse able ins ances include a emp s a
imp o isa ion o explo a ion o code, compa isons wi h p e ious session’s ma e ial
and y o imp o e p o ided code using ial and e o o comple e he ask.
As he lea ning p ocess con inued, he lea ning app oach o he pa icipan s s a ed
o di e . Fo example, Pa icipan 1 (EG) explo ed ou side he scope o he gi en
ask wi h esponses p o ided by ou cha bo as pa o a u he lea ning explo a ion.
Pa icipan 3 (EG) no ed ha he cha bo 's esponse was help ul in s a ing coding,
and a oiding ge ing s uck. Pa icipan 5 (EG) was able o ob ain mo e explana ion
and imp o e hei lea ning s a egy. These examples illus a e he cha bo 's impac
on he pa icipan s' lea ning p ocess, speci ically in helping hem ge uns uck and
ob ain explana ion and cla i ica ion, demons a ing beha iou s ypical o Kembe ’s
e lec ion le el. These indings also sugges ha ques ioning skills a e c i ical o
achie ing he e lec ion le el and sus aining he lea ning p ocess.
C i ical e lec ion is cha ac e ised by a shi in pe spec i e(Kembe e al., 2000;
Leung & Kembe , 2003)(Kembe e al., 2000; Leung & Kembe , 2003). The
pa icipan s om he EG exhibi ed a change in pe spec i e, u ilising he cha bo o
lea ning pu poses and seeking guidance ins ead o simply asking o answe s. Fo
example: Pa icipan 9 (EG) men ioned “I will de ini ely adjus he way I'm using AI.
Yeah, ins ead o gi ing me he answe s di ec ly. Gi e me he s eps and guidance”.
Pa icipan 11(EG) ini ially s uggled and said “ he answe s i (ou cha bo ) ga e was
like no he answe s ha I was looking o ”. In he las session, Pa icipan 11 (EG)
desc ibed ha he example code p o ided by ou cha bo wo ks as e e ence code,
o help o y sol ing he ask “ he cha bo like made me hink mo e deepe han
Cha GPT would, because Cha GPT would like gi e you an answe and his cha bo
would gi e you like a e e ence. So I hink his cha bo was be e o lea ning”.
The CG pa icipan s did no exhibi an obse able change o pe spec i e and ew
ins ances o asking o mo e explana ion o cla i ica ion. The EG demons a ed mo e
ins ances o he highe le el o e lec i e hinking h ough hei abili y o na iga e
lea ning s uggles by asking o explo a ion. This inding aligns wi h p e ious
esea ch sugges ing ha de eloping e lec i e hinking skills equi es suppo i e
ac i i ies (Guo, 2022).
6 CONCLUSIONS
This s udy con ibu es o an inc easing knowledge o how o u ilise gene a i e AI
cha bo s o suppo lea ne s. This s udy indica es ha gene a i e AI has he po en ial
o imp o e s uden s’ e lec i e hinking in lea ning p og amming. The indings
sugges ha a speci ically uned gene a i e AI cha bo acili a es e lec i e hinking.
