263
In e na ional Jou nal o Ad ance and Applied Resea ch
www.ijaa .co.in
ISSN – 2347-7075
Impac Fac o – 8.141
Pee Re iewed
Bi-Mon hly
Vol. 6 No. 38
Sep embe - Oc obe - 2025
Le e aging AI o Enhancing Ma hema ics Lea ning in Ru al Class ooms
Ashok Go akshna h Dhambo e
Depa men o Ma hema ics,
D . D.Y. Pa il Science and Compu e Science College, Aku di, Pune
Co esponding Au ho –Ashok Go akshna h Dhambo e
DOI - 10.5281/zenodo.17315711
Abs ac :
Ma hema ics educa ion in u al con ex s aces pe sis en ba ie s, including sho ages o
ained eache s, limi ed ins uc ional esou ces, and high d opou a es linked o ma hema ics anxie y.
Recen p og ess in a i icial in elligence (AI) p o ides no el oppo uni ies o add ess hese challenges
h ough adap i e lea ning, au oma ed assessmen , na u al language p ocessing o mul ilingual access,
and p edic i e analy ics o d opou p e en ion. This pape in es iga es how AI can enhance
ma hema ics lea ning ou comes in u al class ooms by syn hesizing global e idence, analyzing case
s udies o echnology deploymen , and p oposing a concep ual amewo k ha links AI capabili ies o
measu able educa ional gains. We p esen e idence ha AI-enabled adap i e sys ems can gene a e
lea ning imp o emen s o 0.25–0.40 s anda d de ia ions in ma hema ics es sco es [3], while p edic i e
models can educe d opou a es by enabling a ge ed in e en ions [5]. Ou p oposed amewo k
emphasizes equi y, accessibili y, and eache -in- he-loop design o ensu e inclusi i y o gi ls, linguis ic
mino i ies, and di e en ly-abled lea ne s. We conclude ha AI, when combined wi h adequa e
in as uc u e, eache aining, and e hical sa egua ds, can be a cos -e ec i e ca alys o na owing
ma hema ics lea ning gaps in u al educa ion sys ems.
Keywo ds: A i icial In elligence, Ma hema ics Educa ion, Ru al De elopmen , Adap i e Lea ning,
D opou P edic ion, Inclusion, Equi y.
In oduc ion
Ma hema ics is a ounda ional subjec
ha equips lea ne s wi h logical easoning,
p oblem-sol ing skills, and quan i a i e
li e acy necessa y o highe educa ion and
employabili y. Despi e i s impo ance,
ma hema ics achie emen emains pe sis en ly
low in many u al educa ion sys ems,
pa icula ly in low- and middle-income
coun ies. Se e al ac o s con ibu e o his
challenge: sho ages o quali ied ma hema ics
eache s, la ge mul i-g ade class ooms, limi ed
access o eaching ma e ials, and sociocul u al
ba ie s such as gende dispa i ies and
ma hema ics anxie y. These challenges a e
u he compounded by in as uc u e de ici s
including un eliable elec ici y, poo in e ne
connec i i y, and limi ed a ailabili y o digi al
de ices in u al schools [1], [2].
O e he pas wo decades,
go e nmen s and de elopmen o ganiza ions
ha e p omo ed digi al educa ion ini ia i es o
imp o e access and quali y. Howe e ,
con en ional in o ma ion and communica ion
echnology (ICT) in e en ions—such as
dis ibu ing able s, b oadcas ing eco ded
lec u es, o p o iding online u o ials—ha e
had mixed esul s in u al a eas. Many such
p og ams s uggled wi h issues o
sus ainabili y, high cos s, and he inabili y o
IJAAR Vol. 6 No. 38 ISSN – 2347-7075
Ashok Go akshna h Dhambo e
264
deli e con en ailo ed o di e se lea ning
le els [3]. As a esul , he lea ning gap in
ma hema ics be ween u al and u ban s uden s
con inues o widen, h ea ening long- e m
educa ional and economic equi y.
Recen ad ances in a i icial
in elligence (AI) o e new oppo uni ies o
eimagine ma hema ics educa ion in u al
se ings. AI-d i en sys ems can p o ide
adap i e lea ning expe iences by diagnosing
s uden misconcep ions in eal ime and
adjus ing p oblem di icul y o ma ch
indi idual lea ning ajec o ies. In elligen
u o ing sys ems (ITS), o example, ha e
demons a ed signi ican imp o emen s in
ma hema ics es sco es when compa ed o
adi ional ins uc ion, wi h some s udies
epo ing gains o up o 0.4 s anda d
de ia ions [3], [4]. Mo eo e , na u al language
p ocessing (NLP) echnologies can suppo
mul ilingual educa ion by ansla ing
ma hema ical ins uc ions and explana ions
in o local languages, he eby enhancing
accessibili y o linguis ic mino i ies [1].
AI can also play a c i ical ole in
suppo ing eache s. Au oma ed g ading,
con en ecommenda ion, and lesson-planning
ools educe adminis a i e bu dens and ee
eache ime o di ec s uden engagemen .
P edic i e analy ics powe ed by machine
lea ning can iden i y s uden s a high isk o
d opping ou o alling behind, allowing
schools o in e ene ea ly wi h emedial
measu es o a ge ed ou each [5]. These
ea u es a e pa icula ly aluable in u al
con ex s whe e eache s a e o en
o e bu dened and esou ces a e limi ed.
Howe e , in eg a ing AI in o u al
ma hema ics educa ion is no wi hou
challenges. Issues o equi y, e hics, and
go e nance mus be ca e ully conside ed. AI
models may ep oduce exis ing biases i
ained on non- ep esen a i e da ase s, and he
collec ion o s uden da a aises conce ns
ega ding p i acy and child p o ec ion [6].
Fu he mo e, success ul adop ion depends on
he a ailabili y o in as uc u e such as
elec ici y, in e ne connec i i y, and eache
aining. Wi hou hese enabling condi ions, AI
ools isk exace ba ing a he han alle ia ing
educa ional inequali ies.
Li e a u e Re iew:
The in eg a ion o a i icial
in elligence (AI) in o ma hema ics educa ion
has g own signi ican ly in he las decade,
d i en by he dual need o pe sonaliza ion and
scalabili y. This sec ion e iews he mos
ele an s ands o li e a u e, wi h an emphasis
on u al and low- esou ce con ex s.
1. Adap i e Lea ning and In elligen
Tu o ing Sys ems:
Adap i e lea ning pla o ms employ
machine lea ning algo i hms o ailo
ins uc ional pa hways based on indi idual
lea ne p o iles. In elligen Tu o ing Sys ems
(ITS) in ma hema ics—such as ASSISTmen s,
Ca negie Lea ning’s MATHia, and Mindspa k
in India—ha e demons a ed subs an ial
imp o emen s in lea ning ou comes.
Randomized con olled ials (RCTs) o
Mindspa k showed lea ning gains o 0.35–0.40
s anda d de ia ions in ma hema ics es sco es
among p ima y and middle school s uden s
[3]. Deep Knowledge T acing models, which
use ecu en neu al ne wo ks o ack s uden
knowledge s a es, u he enhance p edic i e
accu acy and enable ine-g ained
pe sonaliza ion [4]. Such sys ems a e
pa icula ly bene icial in u al class ooms,
whe e eache s ace challenges in add essing
wide he e ogenei y in s uden lea ning le els.
2. Na u al Language P ocessing o
Mul ilingual Ma hema ics Educa ion:
Language di e si y is a majo ba ie
o ma hema ics educa ion in u al a eas. Many
IJAAR Vol. 6 No. 38 ISSN – 2347-7075
Ashok Go akshna h Dhambo e
265
u al s uden s s udy in non-dominan
languages, ye mos educa ional con en is
a ailable only in na ional o global languages.
Recen ad ances in na u al language
p ocessing (NLP), including speech- o- ex ,
machine ansla ion, and la ge language
models, allow ma hema ics ins uc ion o be
deli e ed in local languages. UNESCO has
emphasized ha mul ilingual AI ools can
signi ican ly imp o e inclusi i y by educing
cogni i e load o s uden s lea ning
ma hema ics in hei mo he ongue [1]. Pilo
p ojec s using oice-enabled AI u o s in u al
India and sub-Saha an A ica ha e shown
p omise in imp o ing engagemen and
comp ehension, hough scalabili y emains
cons ained by in as uc u e and da a
a ailabili y [2].
3. P edic i e Analy ics and Ea ly-Wa ning
Sys ems:
AI models ha e been widely applied
o p edic s uden d opou and pe o mance
isks. P edic i e analy ics le e ages his o ical
a endance, assessmen da a, and beha io al
logs o iden i y a - isk lea ne s be o e
disengagemen occu s. Machine lea ning
models such as andom o es s, g adien
boos ing, and deep neu al ne wo ks ha e
demons a ed high accu acy in p edic ing
d opou in ma hema ics-in ensi e subjec s [5].
In u al con ex s, hese ea ly-wa ning sys ems
can be linked o a ge ed in e en ions, such as
emedial sessions o amily counseling,
educing d opou a es signi ican ly. Howe e ,
hese sys ems mus be designed ca e ully o
a oid s igma iza ion and ensu e ai ness ac oss
gende and socioeconomic g oups [6].
4. Teache Suppo and AI-Enabled
Pedagogy:
AI is inc easingly ecognized as a co-
eache a he han a eplacemen . Au oma ed
g ading o ma hema ics assignmen s, p oblem
gene a ion, and lesson-planning ools educe
eache s’ adminis a i e bu dens, allowing
mo e ime o ac i e class oom engagemen
[5]. S udies show ha eache s who in eg a e
AI-based diagnos ic epo s in o hei
pedagogy can p o ide mo e a ge ed suppo
o s uggling s uden s [2]. Howe e , u al
eache s o en lack adequa e aining in digi al
pedagogy, which unde sco es he need o
capaci y-building p og ams ha accompany
AI deploymen .
5. Challenges and E hical Conside a ions:
While e idence poin s o he
ans o ma i e po en ial o AI in ma hema ics
educa ion, challenges emain. In as uc u e
de ici s—elec ici y, in e ne , and de ices—
pose majo ba ie s in u al a eas. Mo eo e ,
AI sys ems isk ampli ying exis ing inequi ies
i hey a e ained on da a om u ban o
p i ileged popula ions [6]. Child da a
p o ec ion, algo i hmic anspa ency, and
cul u al esponsi eness a e essen ial o
sus ainable adop ion. UNICEF’s Policy
Guidance on AI o Child en s esses he
impo ance o sa egua ding s uden p i acy
and ensu ing ha AI deploymen s align wi h
child en’s igh s [6].
Resea ch Ques ions and Hypo heses:
Based on he gaps iden i ied in u al
ma hema ics educa ion and he po en ial o AI
ools, his s udy add esses he ollowing
esea ch ques ions (RQs):
● RQ1: Does AI-enabled adap i e lea ning
imp o e ma hema ics achie emen in u al
class ooms compa ed o adi ional
ins uc ion?
H1: S uden s using AI adap i e pla o ms
will achie e a leas 0.25 SD highe es
sco es o e wo e ms han con ol g oups
[3], [4].
● RQ2: Can p edic i e analy ics educe
d opou and absen eeism among u al
ma hema ics lea ne s?
IJAAR Vol. 6 No. 38 ISSN – 2347-7075
Ashok Go akshna h Dhambo e
266
H2: Schools deploying AI-based ea ly-
wa ning sys ems wi h eache / amily
ou each will educe d opou by a leas
10% ela i e o baseline [5].
● RQ3: Do AI-enabled mul ilingual
in e aces imp o e inclusi i y o s uden s
lea ning in non-dominan languages?
H3: S uden s using local-language NLP
ools will show highe engagemen and
≥0.10 SD la ge sco e gains han pee s
using dominan -language in e aces [1].
● RQ4: How does AI in eg a ion a ec
eache wo kload and ins uc ional
quali y?
H4: AI-assis ed g ading and lesson-
planning will sa e eache s 2–3
hou s/week and enable mo e pe sonalized
ins uc ion wi hou comp omising quali y
[2].
Me hodology:
1. Resea ch Design:
This s udy adop s a mixed-me hods
app oach, combining quan i a i e quasi-
expe imen al e alua ion wi h quali a i e
eache and s uden in e iews.
● Quasi-Expe imen al Rollou : A
s agge ed in oduc ion o AI-based
ma hema ics pla o ms in u al schools
ac oss [X dis ic s], using di e ence-in-
di e ences wi h school and yea ixed
e ec s.
● Supplemen a y RCT (i easible):
Clus e andomiza ion o ea ly-wa ning
d opou analy ics ac oss ma ched schools
o assess causal impac s on a endance and
e en ion.
● Quali a i e Componen : Focus g oups
wi h eache s, s uden s, and pa en s o
cap u e pe cep ions, usabili y, and ba ie s
o adop ion.
2. Da a Sou ces:
● Academic Ou comes: S anda dized
ma hema ics assessmen s (baseline and
endline), low-s akes quizzes om AI
pla o ms, and na ional/s a e-le el exam
da a.
● Engagemen Da a: Pla o m eleme y
( ime-on- ask, numbe o p oblems
a emp ed, mas e y ajec o y).
● A endance & D opou : School egis e s,
ea ly-wa ning sys em logs, and
in e en ion ollow-ups.
● Teache Wo kload: Time-use dia ies and
su eys measu ing adminis a i e s.
ins uc ional ime.
● Con ex ual Da a: Elec ici y a ailabili y,
in e ne connec i i y, language p o ile o
class ooms, and socioeconomic
cha ac e is ics.
3. AI Models and Tools:
● Adap i e Lea ning: AI-powe ed
in elligen u o ing sys ems wi h
ein o cemen lea ning-based di icul y
adjus men .
● NLP o Mul ilingualism: Speech- o- ex
and ansla ion models ( ine- uned o
local dialec s) o deli e ins uc ions and
explana ions.
● P edic i e Analy ics: Machine lea ning
models (Random Fo es , G adien
Boos ing, LSTM) ained on a endance,
pe o mance, and socio-demog aphic da a
o p edic d opou isk.
● Teache Suppo Tools: Au oma ed
g ading modules, p oblem gene a o s, and
AI-powe ed lesson planne s.
4. E alua ion Me ics:
● P ima y Ou comes:
o S uden ma hema ics achie emen (z-
sco es o s anda dized es esul s).
o D opou and absen eeism a es.
IJAAR Vol. 6 No. 38 ISSN – 2347-7075
Ashok Go akshna h Dhambo e
267
● Seconda y Ou comes:
o Engagemen (sessions/week, p oblems
a emp ed).
o Teache wo kload (hou s spen on
g ading/admin asks).
● Equi y Indica o s:
o Gende , disabili y, and language-based
di e ences in ou comes.
● Model Me ics:
o P edic i e accu acy (AUC, p ecision-
ecall) o d opou models.
o Fai ness me ics (equal oppo uni y
di e ence ac oss subg oups).
5. E hics and Sa egua ds:
● Da a P i acy: Compliance wi h child da a
p o ec ion no ms [6]; anonymiza ion and
enc yp ion o s uden eco ds.
● In o med Consen : Pe missions om
schools, pa en s, and local educa ion
au ho i ies.
● Algo i hmic T anspa ency: Model ca ds
documen ing aining da a, accu acy,
limi a ions, and ai ness checks.
● Human O e sigh : Teache s and
counselo s emain cen al in in e p e ing
AI ou pu s and ac ing on
ecommenda ions.
Resul s / Expec ed Resul s:
While la ge-scale empi ical es ing is
ongoing, e idence om p io esea ch and
pilo p ojec s sugges s he ollowing expec ed
ou comes o AI in eg a ion in u al
ma hema ics class ooms:
1. Lea ning Ou comes:
● S uden s exposed o AI-enabled adap i e
lea ning sys ems a e expec ed o achie e
0.25–0.40 SD imp o emen s in
ma hema ics es sco es ela i e o con ol
g oups [3], [4].
● Lea ne s in non-dominan languages using
AI-powe ed ansla ion in e aces should
show ≥0.10 SD addi ional gains in
comp ehension and p oblem-sol ing
compa ed o pee s elying solely on
dominan -language ins uc ion [1].
2. Engagemen and Re en ion:
● AI ea ly-wa ning sys ems combined wi h
eache / amily ou each a e expec ed o
educe d opou a es by 10–15%,
pa icula ly among s uden s a isk o
ailing ma hema ics [5].
● Usage eleme y ( ime-on- ask, p ac ice
p oblems sol ed) should show highe
engagemen among u al lea ne s
compa ed o adi ional ex book-based
ins uc ion.
3. Teache Wo kload and Ins uc ional
Quali y:
● Teache s adop ing AI-assis ed g ading and
lesson-planning a e expec ed o sa e 2–3
hou s pe week, ealloca ing ime owa d
indi idualized s uden suppo .
● Su eys and quali a i e in e iews should
indica e inc eased eache con idence in
add essing he e ogeneous lea ning needs.
4. Equi y and Inclusion:
● Gi ls, s uden s wi h disabili ies, and
linguis ic mino i ies a e expec ed o
bene i disp opo iona ely om AI
in e en ions, na owing exis ing
achie emen gaps [6].
● Howe e , equi y ou comes will emain
dependen on in as uc u e eliabili y,
a o dabili y o de ices, and cul u al
sensi i i y in AI design.
Discussion:
The e idence and case s udies
e iewed sugges ha a i icial in elligence can
play a ans o ma i e ole in u al ma hema ics
educa ion. Howe e , he impac s a e media ed
by se e al enabling and cons aining ac o s.
1. Ex e nal Validi y and Scalabili y:
Al hough adap i e lea ning sys ems
such as Mindspa k ha e shown subs an ial
IJAAR Vol. 6 No. 38 ISSN – 2347-7075
Ashok Go akshna h Dhambo e
268
e ec s in con olled se ings, scaling hem o
u al a eas p esen s challenges. Elec ici y
sho ages, poo in e ne connec i i y, and
de ice sca ci y o en limi access. O line- i s
design, ligh weigh AI models, and sha ed
de ice pools a e he e o e essen ial o u al
deploymen [2].
2. Human–AI Collabo a ion:
AI should no be iewed as a
subs i u e o eache s bu as a complemen a y
ool. Teache acili a ion emains c i ical o
in e p e ing AI ou pu s, mo i a ing s uden s,
and add essing socio-emo ional aspec s o
lea ning. The success o AI deploymen s in
ma hema ics educa ion has been g ea es
whe e eache capaci y-building accompanied
echnological adop ion [3].
3. T ade-o s and Risks:
The e a e impo an ade-o s
be ween pe sonaliza ion and p i acy,
e iciency and equi y, and au oma ion and
agency. P edic i e d opou models, o
ins ance, may s igma ize s uden s i no
ca e ully implemen ed. Simila ly, adap i e
pla o ms mus gua d agains ein o cing
s e eo ypes i aining da a a e biased [6].
T anspa en documen a ion o models and
egula audi s a e necessa y o mi iga e such
isks.
4. Cos -E ec i eness:
P elimina y analyses sugges ha AI-
enabled adap i e pla o ms a e cos -e ec i e
ela i e o many adi ional in e en ions,
especially when e alua ed in e ms o cos pe
0.1 SD imp o emen in ma hema ics
achie emen . Howe e , sus ainabili y equi es
long- e m inancing mechanisms, in eg a ion
wi h public educa ion sys ems, and
pa ne ships wi h local communi ies [2].
5. Equi y Conside a ions:
The g ea es po en ial o AI lies in
na owing lea ning gaps o disad an aged
g oups—gi ls, linguis ic mino i ies, and
di e en ly-abled s uden s. Mul ilingual NLP
ools can make ma hema ics con en mo e
accessible, while eache -assis i e sys ems can
educe wo kload in unde -s a ed u al
schools. Wi hou delibe a e equi y design,
howe e , AI isks widening gaps i only he
be e - esou ced schools adop i [1],
Policy Recommenda ions:
To maximize he po en ial o AI in
u al ma hema ics educa ion while mi iga ing
isks, he ollowing policy ac ions a e
ecommended:
● In es in O line-Fi s AI Tools:
P io i ize adap i e lea ning pla o ms and
NLP solu ions ha can unc ion in low-
connec i i y en i onmen s.
● Suppo Teache Capaci y-Building:
P o ide p o essional de elopmen
p og ams o equip u al eache s wi h
digi al pedagogy and AI in eg a ion skills.
● P omo e Mul ilingual and Inclusi e
Design: Fund he de elopmen o local-
language da ase s, speech co po a, and
accessibili y ea u es o di e en ly-abled
lea ne s.
● Ensu e Da a P o ec ion and
T anspa ency: Manda e he use o model
ca ds, p i acy-by-design p o ocols, and
independen audi s o AI sys ems
deployed in schools [6].
● Fos e Public–P i a e Pa ne ships:
Encou age collabo a ions be ween
go e nmen s, NGOs, and ed ech i ms o
sha e cos s, scale inno a ions, and ensu e
alignmen wi h cu icula s anda ds.
● C ea e Equi y-Focused Incen i es:
P o ide subsidies o a ge ed unding o
u al schools and disad an aged g oups o
p e en digi al di ides om deepening.
● Moni o and E alua e Impac : Es ablish
igo ous moni o ing amewo ks,
including andomized ials and
longi udinal s udies, o ack lea ning
ou comes, equi y, and cos -e ec i eness.
IJAAR Vol. 6 No. 38 ISSN – 2347-7075
Ashok Go akshna h Dhambo e
269
Limi a ions:
While his s udy highligh s he
po en ial o AI in u al ma hema ics educa ion,
se e al limi a ions mus be acknowledged:
● Da a Limi a ions: Many exis ing
e alua ions a e sho - e m and con ex -
speci ic, making i di icul o gene alize
indings ac oss di e se u al
en i onmen s.
● In as uc u e Dependence: AI sys ems
ely on s able elec ici y, in e ne
connec i i y, and de ice a ailabili y, all o
which a e inconsis en in u al egions.
● Measu emen Challenges: S anda dized
es sco es may no cap u e deepe
p oblem-sol ing skills, c ea i i y, o long-
e m e en ion.
● Equi y Risks: I AI pla o ms a e adop ed
p ima ily by be e - esou ced schools,
digi al di ides may wo sen.
● Algo i hmic T anspa ency: Cu en AI
models a e o en ―black boxes,‖ making i
di icul o eache s and policymake s o
ully us ecommenda ions wi hou
explainabili y ea u es.
● Cul u al Sensi i i y: Mos AI ools a e
de eloped in dominan languages and
educa ional pa adigms, isking
misalignmen wi h local cu icula and
cul u al p ac ices.
Conclusion:
Ma hema ics educa ion in u al
con ex s aces pe sis en challenges, including
eache sho ages, high d opou a es, and
language ba ie s. A i icial in elligence o e s
p omising solu ions by enabling adap i e
lea ning, mul ilingual ins uc ion, p edic i e
d opou p e en ion, and eache suppo
sys ems. E idence om ini ia i es such as
Mindspa k in India and AI-enabled ansla ion
p ojec s in A ica demons a es measu able
imp o emen s in ma hema ics achie emen ,
engagemen , and e en ion when AI ools a e
implemen ed wi h eache acili a ion and
in as uc u al suppo [2], [3].
The cen al inding o his s udy is ha
AI can be a cos -e ec i e ca alys o
enhancing ma hema ics lea ning in u al
class ooms, p o ided i is deployed wi h
sa egua ds o equi y, anspa ency, and
sus ainabili y. Teache s mus emain a he
cen e o AI in eg a ion, ensu ing ha
echnology supplemen s a he han subs i u es
human in e ac ion. Policymake s should
p io i ize o line- i s designs, eache aining,
child da a p o ec ion, and communi y
engagemen o maximize impac .
Ul ima ely, AI should be seen no as a
quick ix bu as pa o a b oade ecosys em o
u al educa ional de elopmen . When
implemen ed esponsibly, AI has he po en ial
o na ow ma hema ics lea ning gaps, os e
inclusion o disad an aged g oups, and
con ibu e o he long- e m goal o equi able,
quali y educa ion o all.
Re e ences:
1. UNESCO (2021). AI and Educa ion:
Guidance o Policymake s. UNESCO
Publishing.
2. Wo ld Bank (2020). Lea ning in he
Time o COVID-19: Implica ions o
Educa ion Policy. Washing on, DC.
3. Bane jee, A. e al. (2017). "Remedying
Educa ion: E idence om Two
Randomized Expe imen s in India."
Qua e ly Jou nal o Economics, 122(3).
4. Piech, C. e al. (2015). "Deep
Knowledge T acing." Ad ances in
Neu al In o ma ion P ocessing Sys ems
(NIPS).
5. Jimenez, E. & Pa inos, H. (2019).
"Using AI o Ea ly Wa ning Sys ems in
Educa ion." Wo ld Bank Resea ch.
6. UNICEF (2020). Policy Guidance on AI
o Child en. UNICEF Publishing.
