223

CHAPTER 15

INTELLIGENT TUTORING SYSTEMS FOR ENHANCING

ACADEMIC PERFORMANCE OF SECONDARY STUDENTS

IN INDIA

Dr. Rimmi Datta

Abstract

Background:

significant portion of classrooms have large within-grade learning dispersion, which limits teacher-centered instruction

and leads to poor academic performance.

Objective:

performance in secondary-stage students in India, a combination of national official data and a multiple-case case

study of published PAL/ITS applications.

Methods:

Explanatory multiple-case design that is desk-based and has embedded units (school contexts). The

researcher triangulated: (i) national statistics and assessments (UDISE+, ASER 2024, National Council of

Educational Research and Training large-scale assessment systems); (ii) peer-reviewed and working-paper RCTs

learning gains can be obtained on independent tests (math: 0.22-0.43 SD at scale; 0.37 SD in an efficacy trial)

and grade level school tests might fail to reflect such gains when instruction aims several years below grade norms.

Conclusions:

under realistic conditions of the public-system, but the improvement in performance requires dosage, device access,

teacher integration, governance and redesigning of assessment. Existing ICT funding under Samagra Shiksha must

be the policy pathways, in line with NDEAR interoperability, and meet child-data protection requirements in the

then-developing data protection laws.

224

Keywords:

Intelligent Tutoring Systems; Personalized Adaptive Learning; Learning Outcomes; Randomized

Controlled Trial; Digital Infrastructure.

Introduction

in post-primary grades, a significant fraction of students are several years below the norms of the

curriculum, suggesting that coverage of syllabus can not necessarily mean that students have

mastered the pre-requisite skills, particularly in mathematics and language. [2,11] [19]

Why ITS now? ITS/PAL is especially timely in India in 2026, as there are two conditions. To

begin with, digital accessibility among schools is on the rise: the number of schools that report

having access to computers has expanded to 64.7% and access to internet to 63.5% in 202425

(compared to 57.3% and 53.6% in 202325). Second, the national digital educational ecosystem is

converging to interoperable platforms and reusable building blocks (e.g., NDEAR; DIKSHA),

allowing assessment services, identity services, content services, analytics services, etc., to be

similarly modularly integrated. [20]

3) What policy design features—financing, interoperability, assessment alignment, and data

governance—are required for durable gains in India’s public secondary education system?

Assumptions and scope: Since no particular state/region/site was given, the researcher (i)

considers this to be a national-level synthesis using official, all-India indicators; and (ii) uses two

exemplary school-level cases using documented pilots an urban after-school model (Delhi

catchment) and a rural-heavy in-school model (Rajasthan Adarsh schools) supplemented with a

225

government-led scale model (Andh These include mostly Classes 6 9; we explain our relevance to

secondary education by (a) Class 9 is in secondary and (b) the accumulated deficit of skills in Class

8 9 has a material effect on secondary academic achievement and readiness to pass board exams.

Literature review and theoretical framework

the attempt of a learner to solve items, first mathematically modeled by Albert T. Corbett and John

R. Anderson in Bayesian form. [18]

Effectiveness evidence: what the global literature says? Meta-analytic research in

educational psychology shows that ITS tend to be effective, although the sizes of effects differ

depending on outcome measure, comparison condition and context. Wenting Ma and colleagues

combined 107 effect sizes (14,321 participants) and found positive effects in education levels and

areas. According to James A. Kulik and J. D. Fletcher, a median effect of about 0.66 SD was

observed in 50 controlled ITS assessments, and they point out that measured gains were strongly

dependent on the fit between assessment and instructional objectives- a problem that lies at the

heart of the Indian scale cases discussed here. Kurt VanLehn also defines families of design based

students of various grades of achievement in one classroom. As a result of this heterogeneity, the

efficacy of one-pace instruction is diminished and focused remediation is a realistic high-leverage

intervention.

Theoretical framework: mastery learning and teaching at the right level: The mastery

learning and curricular-right-sizing are integrated in the theoretical lens. Mastery learning,

popularised by Benjamin S. Bloom[36], argues that the majority of learners can attain high levels

226

of mastery with adequate time, good feedback and corrective instructions that is, the processes of

diagnostics and personalization are not peripheral tools. PAL/ITS puts this mechanism into

practice: they are diagnosing followed by providing customized practice with instant feedback. The

India PAL evidence base is consistent with this theory: massive gains are seen in independent

adaptive tests despite small change seen in grade-level tests, which are in line with learning recovery

on below-grade baselines.

Policy and system alignment: The policy architecture in India also favors the fairness of

through to Class X; this provides a financing mechanism through which ITS/PAL labs can be

funded. [4] Interoperability aspirations in NDEAR also mean that ITS/PAL must be viewed as

being modular services that are combined with national platforms and not pilots. [5]

Methodology

Design: The study design implemented by the researcher was an explanatory and embedded

multiple-case study (desk-based) which is appropriate in situations where (i) causal findings are

obtained by the rigorous quantitative appraisal, but (ii) the uptake of the policy would require

tablets), and the governance model (NGO supported vs state-run).

 Case A (Urban efficacy): Delhi after-school PAL/ITS program (Mindspark) was the

strategy tested through a scholarship/lottery allocation system among learners in Grades 69 among

public middle schools. [10]

 Case B (Scale adaptation): The Rajasthan in-school Mindspark program was part of the

227

Embedded units were “school implementation contexts” (urban sample, rural-heavy sample,

and statewide government lab model). In cases where school-specifics were not publicly listed, we

documented only features that were available and made clear what had been assumed in operation

(e.g. timetabling norms), without fabricating it.

Data sources: The researcher triangulated four source classes:

1) Official national statistics and infrastructure indicators from UDISE+ 2024–25 key

findings. [1] [2]

2) Learning and enrolment benchmarks from Annual Status of Education Report 2024[46]

(ASER 2024). [2]

3) National assessment system documentation from NCERT/PARAKH resources (NAS

measurement instruments (independent assessments vs school exams; item ranges; standardization

method); (iv) implementation model (hardware, staffing, teacher role, monitoring); and (v)

outcomes (effect sizes, usage, subgroup heterogeneity, exam impacts, cost parameters). Threats to

scale (dose reduction, displacement of instructional time, teacher adaptation) were also extracted

by the researcher.

Analysis strategy: The within-case synthesis generated logic models between implementation

features and learning outcomes; the cross-case synthesis compared the effect sizes and

mechanisms. The researcher report standardized treatment effects (SD units) as the key similar

measure, noting that tests in different cases are not identical; therefore, cross-case comparisons

are interpretive, but not mechanistically the same (assumption mentioned). The researcher also

report conversions of equivalent years of schooling used by authors of studies where it is available.

[17]

228

229

230

and DIKSHA (foundational national platform that will allow distribution, but is not an ITS). The

main policy implication is that the ecosystem in India most probably needs: (i) an interoperable

rail (DIKSHA/NDEAR-aligned services), and (ii) PAL/ITS-based applications (evidence-based)

that are interconnected into rails. [3,5]

231

(i) in cases where tests are adaptive or measure a broad range of abilities, the measured effects are

large; (ii) in cases where the measure of performance is based on grade level school exams, the

effects of learning may be insignificant even when learning is higher, because instruction is below

grade-level. [17]

(Grades 1–12),

across rural

and urban

areas

annually; key

grades analyzed

include Grade 5

and Grade 8 [11]

C: Andhra

Pradesh

government-

run PAL labs

Tablet-

based PAL

labs; 2×40-

minute weekly

Grades 6–9

across 120

schools in eight

districts (60

Randomized

Learning gains: cross-case visualization: The standardized learning gains in mathematics in

the three cases in India are visualized in figure 3 (with an accent on the fact that) (a) large short-

run efficacy gains can still decline at scale (usually due to a reduction in dosage), and (b) large

impacts can also be achieved with well-manageable scale models.

Figure 3