126

CHAPTER 8

GENERATİVE AI (GENAI) İN SCİENCE EDUCATİON AS

AN INNOVATİVE PRACTİCE: A SYSTEMATİC REVİEW

Dr. Yashpal D. Netragaonkar

Associate Professor, Department of Education

Dr. Vishwanath Karad, MIT World Peace University, Pune-38, India Email:

dryashdnet@gmail.com Mobile: 9881595917

ORCID: 0009-0002-2035-7421

Abstract

Generative artificial intelligence (GenAI)—particularly large language model (LLM) tools—has rapidly entered

educational practice and is beginning to reshape science teaching, learning, and assessment. Science education is a

distinctive use case because it requires epistemic reliability: learners must justify claims with evidence, apply

disciplinary constraints (e.g., units, conservation laws), and engage in inquiry practices. This chapter offers a

PRISMA 2020–aligned systematic review of recent research on GenAI in science education, complemented with a

policy and ecosystem analysis for India. Evidence from published systematic reviews and empirical studies suggests

that GenAI can support explanation, scientific writing, formative feedback, and inquiry planning when embedded

in well-designed tasks. However, risks persist: hallucinations and inaccuracies, bias, privacy concerns, and academic

integrity threats, especially where institutional guidance is limited. In India, NEP 2020 and NCF-SE 2023

emphasize competency-based learning, technology integration, and scientific temper, while national digital

infrastructure (DIKSHA and NDEAR) provides a scalable platform for teacher professional development and

content delivery. This chapter synthesizes evidence into an India-ready implementation framework

‑

CIENTIFIC), proposes assessment redesign options, and provides classroom-ready prompt templates, rubrics,

and a reproducible search strategy (databases and Boolean strings).

Keywords:

Generative AI, Large Language Models, Science Education, Systematic Review, İnquiry Learning,

AI literacy, NEP 2020, NCF-SE 2023, DIKSHA, NDEAR, IndiaAI Mission.

Introduction

Science education aims to cultivate scientific temper, conceptual understanding, evidence-based

reasoning, and the ability to investigate phenomena. These aims require learners to move beyond

memorization toward explanation, modeling, and argumentation. In this landscape, GenAI has

emerged as a disruptive yet potentially empowering innovation. LLM-based tools can generate

natural-language explanations, questions, summaries, and feedback; they can also help learners

draft laboratory reports, compare alternative models, generate code for basic data analysis, and

translate technical language into accessible forms. Systematic reviews of ChatGPT and related

127

GenAI tools in education report perceived benefits such as immediate feedback, personalization,

and improved access to learning support, especially for writing-intensive tasks (Bettayeb & Abu

Talib, 2024).

At the same time, science education is a demanding domain for GenAI because scientific

knowledge is constrained by evidence, measurement, and formal principles. GenAI outputs may

be fluent yet incorrect (hallucinations’), potentially reinforcing misconceptions if used without

verification routines. Systematic reviews of GenAI in pedagogical practices highlight recurring

concerns: inaccuracies, bias, threats to academic integrity, and uncertainty about how GenAI

changes learners’ cognitive effort and metacognition (Wang et al., 2025). These concerns have

motivated international guidance calling for human-centered, safe, equitable, and age-appropriate

adoption of GenAI in education (UNESCO, 2023).

The Indian education context creates both opportunities and constraints for GenAI adoption.

NEP 2020 emphasizes technology integration and explicitly states that technology interventions

should be rigorously and transparently evaluated in relevant contexts before scaling (Ministry of

Education, 2020). India also has large-scale digital education infrastructure through DIKSHA, a

national platform offering curriculum-aligned digital content and teacher professional

development across languages (DIKSHA, 2026; Digital India, 2026). Complementing DIKSHA,

the National Digital Education Architecture (NDEAR) provides a unifying, interoperable

framework to connect education services and platforms while emphasizing privacy and security by

design (Ministry of Education, 2022).

The National Curriculum Framework for School Education (NCF-SE) 2023 operationalizes

NEP 2020’s competency-based vision and includes subject guidance for science education across

stages (Ministry of Education, 2023). In the school sector, CBSE has introduced Artificial Intelligence

as a skill subject (e.g., Subject Code 417) and developed manuals for AI integration across subjects,

including science (CBSE, 2024a; CBSE, 2020). These developments indicate that India is building

curricular and infrastructural readiness for AI-related learning. However, learning about AI’ (AI

literacy and skills) is distinct from ‘learning with GenAI’ (using LLM tools to support learning in

other subjects).

Given the rapid diffusion of GenAI tools outside institutional control, science educators and

policymakers face urgent questions: What does research evidence say about GenAI’s impact on

science learning? Which classroom uses are beneficial, which are risky, and under what conditions?

How can Indian schools and HEIs adopt GenAI in ways that align with NEP 2020 and NCF-SE

2023 while protecting privacy, equity, and academic integrity? This chapter addresses these

questions through a systematic review and an India-focused synthesis.

128

Conceptual Background: Why Science Education is a Distinctive GenAI Use Case

GenAI tools are general-purpose: they can generate coherent text and respond conversationally

across topics. In science education, however, quality is not simply readability; it is epistemic

warrant. Learners must justify claims with data, apply constraints (units, conservation laws,

boundary conditions), and evaluate alternative explanations. Therefore, GenAI can be

educationally powerful only when it is embedded in tasks that preserve learner agency and require

verification. UNESCO’s guidance emphasizes ethical validation, protection of privacy, and

human-centered pedagogical design (UNESCO, 2023).

From a learning sciences perspective, GenAI can be positioned as (a) a cognitive scaffold that

prompts explanation, reflection, and revision; (b) a ‘second voice’ that offers alternative hypotheses

and representations; or (c) an automated answer generator that may reduce productive struggle.

The literature suggests that outcomes depend strongly on task design and teacher mediation.

Reviews in pedagogy report benefits when GenAI is used as a supplementary tool for feedback

and idea generation but warn against overreliance and reduced critical thinking if students

outsource reasoning (Wang et al., 2025).

In India, science education priorities include developing scientific temper, inquiry, and

application of knowledge to local and national challenges. NCF-SE 2023 frames science learning

as building process skills such as observation, analysis, inference, and evidence-based thinking,

aligning with broader aims of NEP 2020 (Ministry of Education, 2023). Responsible GenAI

integration can support these goals—for example, by generating prompts for data interpretation,

proposing alternative models to critique, or helping students express scientific reasoning in their

home language. However, equitable access and language performance differences must be

considered to avoid widening learning gaps.

Methods: PRISMA 2020–Aligned Systematic Review Approach

This chapter follows PRISMA 2020 reporting principles to transparently describe the review

purpose, methods, and synthesis approach (Page et al., 2021). Because research on GenAI in

science education is recent and heterogeneous, the review uses a narrative thematic synthesis rather

than a quantitative meta-analysis. The review is complemented by a policy and ecosystem scan for

India (NEP 2020, NCF-SE 2023, DIKSHA, NDEAR, CBSE AI initiatives, and the IndiaAI Mission).

Research questions guiding the review were: (RQ1) What are the dominant GenAI use cases in

science education (teaching, learning, assessment, inquiry)? (RQ2) What outcomes are reported

(learning, engagement, scientific writing, reasoning quality)? (RQ3) What risks and challenges are

129

documented (accuracy, bias, integrity, privacy, equity)? (RQ4) What implementation conditions

enable responsible, effective use (AI literacy, prompt design, verification routines, governance)?

Search strategy overview. A reproducible search strategy (databases and Boolean strings) is

provided in Addendum C. Recommended databases for peer-reviewed studies include Scopus,

Web of Science, ERIC, and Google Scholar, with optional inclusion of IEEE Xplore/ACM for

STEM intersections. Policy sources include UNESCO, OECD, and Indian education

policy/curriculum documents (UNESCO, 2023; OECD, 2023; Ministry of Education, 2020; Ministry

of Education, 2023).

Eligibility criteria. Included works were: (a) peer-reviewed empirical studies, design-based

research, or systematic reviews; (b) studies involving GenAI/LLM tools used for

teaching/learning/assessment in science or STEM; (c) studies reporting outcomes, user

perceptions, or implementation insights. Excluded works were: purely technical model papers

without educational context; opinion pieces without methods; and non-education applications.

Synthesis. Included studies were coded by education level, science discipline, GenAI task type

(explanation, writing, inquiry, assessment), reported outcomes, risks, mitigation strategies, and

contextual factors (policy, access, training). Themes were developed iteratively and reported as a

narrative synthesis.

PRISMA 2020 FLOW DIAGRAM

Figure 1. PRISMA 2020 flow diagram

130

Results: Thematic Synthesis of Evidence on GenAI in Science Education

GenAI for explanation, concept clarification, and tutoring

A dominant use case is employing GenAI as an on-demand explanation partner. Learners ask

conceptual questions (e.g., chemical equilibrium, Newton’s laws, genetics) and receive

conversational explanations, examples, and analogies. Systematic reviews report benefits such as

rapid access to information, personalized responses, and improved learning support—particularly

for learners seeking clarification outside classroom time (Bettayeb & Abu Talib, 2024).

However, GenAI responses can be inaccurate or overconfident. In science education, errors

may involve incorrect causal mechanisms, misapplied formulas, or misunderstandings of

experimental design. Effective practice requires a ‘verification layer’: students compare GenAI

responses against textbooks, teacher notes, simulations, or laboratory data. UNESCO’s guidance

recommends ethical validation and human supervision to ensure safe, meaningful use (UNESCO,

2023).

GenAI for scientific writing, lab reports, and communication

GenAI is frequently used to support scientific writing: organizing lab reports, refining grammar,

and providing formative feedback. Reviews of GenAI in pedagogy report improved instructional

efficiency through faster feedback and personalized materials, alongside perceived gains in

engagement (Wang et al., 2025). For science education, writing support is beneficial when it helps

students express reasoning clearly, but becomes problematic when GenAI replaces the student’s

scientific thinking.

A practical distinction is between language-level assistance (clarity, structure) and reasoning-

level outsourcing (inventing results, fabricating interpretations). Reviews highlight academic

integrity concerns (Bettayeb & Abu Talib, 2024). Process-oriented assessment designs—raw data

submission, drafts, GenAI logs, and oral defence—help preserve authenticity while still leveraging

GenAI for revision.

GenAI for inquiry: hypothesis generation and experimental planning

Emerging work explores GenAI for inquiry-based science learning: brainstorming hypotheses,

identifying variables, planning procedures, and anticipating sources of error. Syntheses suggest

GenAI can catalyze idea generation and support problem-solving when used as a guided tool (Wang

et al., 2025). The highest value often comes from prompting learners to consider alternatives, justify

choices, and identify confounds, rather than producing a single ‘best’ answer.

131

In India, inquiry tasks can be strengthened by contextualizing science in local phenomena (e.g.,

water quality, heat waves, air pollution, agriculture, biodiversity). GenAI can help teachers generate

locally relevant question sets and data-collection templates, while students still conduct

observation and measurement. Safety remains essential: experimentation must stay within teacher-

approved, age-appropriate protocols.

Assessment pressures: integrity, authenticity, and redesign

Assessment is consistently identified as a pressure point. Reviews report concerns that students

may submit AI-generated work as their own, undermining authenticity and fairness (Bettayeb &

Abu Talib, 2024; Wang et al., 2025). In response, educators are shifting toward assessment designs

that emphasize reasoning processes, data interpretation, and oral explanation—outcomes that are

more difficult to outsource.

OECD analysis emphasizes the need for trustworthy and equitable digital ecosystems and

guardrails around AI use (OECD, 2023). In India, assessment redesign aligns with competency-

based approaches in NCF-SE 2023, which emphasizes learning outcomes and process skills

(Ministry of Education, 2023). Examples include in-class data analysis tasks, viva voce, lab practicals,

and iterative projects requiring evidence logs and reflection.

Figure 2. Conceptual matrix of assessment robustness to unattributed GenAI use (design-dependent;

illustrative).

132

AI literacy as a mediator of benefits and harms

Across the literature, AI literacy—understanding what GenAI can and cannot do—emerges as

a core mediator. Educational impact depends on instructor guidance, institutional policies, and

students’ capacity to critically evaluate AI outputs (Bettayeb & Abu Talib, 2024). In science education,

epistemic AI literacy’ is especially important: students must ask what counts as evidence, what

assumptions are present, and how claims could be tested or falsified.

This aligns with NEP 2020’s emphasis on critical thinking and ethical awareness around

emerging technologies (Ministry of Education, 2020) and UNESCO’s human-centered vision

(UNESCO, 2023).

Multimodal GenAI and representation translation

GenAI systems are increasingly multimodal, enabling interaction across text, images, and code.

In science education, this can support translation among representations—verbal explanation,

equations, graphs, and diagrams. Yet multimodal outputs can also embed errors (e.g., wrong axis

interpretation, misleading diagrams). Therefore, teachers should explicitly teach checking routines:

unit checks, dimensional analysis, constraint checking against physical laws, and comparison with

verified sources.

Teacher workload and professional practice

Teachers use GenAI for lesson planning, worksheet generation, and differentiation. Potential

efficiency gains are reported, but generated materials must be validated for curricular alignment

and scientific accuracy (Wang et al., 2025). In India, DIKSHA can support validation by offering

curriculum-aligned resources for cross-checking and by hosting teacher professional development

modules on responsible GenAI use (DIKSHA, 2026).

Indian Policy, Curriculum, and Digital Ecosystem

India’s policy and infrastructure environment provides several enablers for responsible GenAI

adoption in science education. NEP 2020 encourages technology integration while calling for

careful evaluation and attention to privacy and ethics (Ministry of Education, 2020). NCF-SE 2023

operationalizes competency-based science learning and recognizes ICT as cross-cutting (Ministry of

Education, 2023). DIKSHA provides an at-scale repository of digital resources and teacher

professional development (DIKSHA, 2026). NDEAR provides an interoperable architecture with

privacy and security by design (Ministry of Education, 2022). Together, these enable an approach

where GenAI is integrated through guided pedagogy and verification resources rather than ad hoc,

unsupervised use.

133

CBSE and AI readiness as a bridge to GenAI literacy

CBSE’s AI curriculum introduces AI readiness, the AI project cycle, basic Python, and ethical

considerations such as bias and access (CBSE, 2024). This ‘learning about AI’ pathway can be used

to support ‘learning with GenAI’ by making students aware of model limitations and responsible-

use expectations.

IndiaAI Mission and indigenous capacity

India AI Mission aims to strengthen India’s AI ecosystem through compute capacity, datasets,

innovation, applications, future skills, startup financing, and safe and trusted AI (Press Information

Bureau, 2024; IndiaAI, 2026). For education, these pillars can support indigenous, multilingual

models and governance tools that align better with India’s linguistic diversity and curricular

priorities, while also enabling teacher training and safe deployment pathways.

Practice Framework for Indian Science Education

Figure 3. S

‑

CIENTIFIC framework for responsible GenAI integration in science education (proposed in this

chapter).

S‑CIENTIFIC translates research themes and policy principles into design commitments: (S)

Source-and-scope constraints; (C) Claim–Evidence–Reasoning rewriting; (I) Inquiry prompts; (E)

Explain edits transparently; (N) No-private-data rule; (T) Teach epistemic vigilance; (I) Integrity-

by-design assessment; (F) Feedback loops for teachers; (I) Inclusion and access planning; (C)

Continuous policy alignment.

134

Integrating GenAI into the 5E inquiry cycle

Figure 4. GenAI-supported 5E inquiry cycle

A key principle is to use GenAI to amplify inquiry rather than to shortcut it. In Engage, GenAI

can help generate curiosity questions connected to local contexts. In Explore, it can propose

variable lists and data-table formats while the teacher enforces safety and feasibility. In Explain,

GenAI can help students rewrite explanations into CER form, but students must cite evidence and

verified sources. In Elaborate, GenAI can propose alternative models to critique and link concepts

to SDGs. In Evaluate, teachers can use GenAI to draft viva questions and feedback prompts,

while the teacher remains the final assessor.

Low-bandwidth and multilingual implementation options

Because internet and device access vary, equity-first implementation is essential. Options

include: (a) teacher-mediated whole-class GenAI use via projector, (b) rotational station use in

computer labs, (c) offline-first verification anchored to DIKSHA resources, and (d) bilingual

scaffolding to support comprehension while maintaining scientific precision.

Assessment of redesign options (integrity-by-design)

Assessment redesign can reduce incentives and opportunities for unattributed GenAI use.

Options include in-class data tasks, oral defence/viva, lab practicals with observation checks,

135

iterative drafts with process evidence, and reflective prompts that require students to justify

decisions. These approaches align with competency-based assessment principles in NCF-SE 2023

and address integrity risks highlighted in the literature.

Tables for Classroom and Policy Implementation

Table 1. GenAI application patterns aligned to Indian curriculum priorities

GenAI

use case

Example

science task

Science

practice

supported

Key risk Indian

alignment

Concept

clarification

Explain

diffusion vs

osmosis with

local example;

identify

misconceptions

Explanation

, conceptual

change

Hallucinations;

oversimplification

NCF-SE

outcomes; NEP

2020 critical

thinking

CER

writing

support

Rewrite lab

conclusion into

CER; improve

clarity

Scientific

communicatio

Reasoning

outsourcing

Competency

-based

assessment

Inquiry

planning

Plan

variables/control

s for filtration

experiment

Inquiry,

design

Unsafe/impractica

l procedures

Experiential

learning

emphasis

Formative

feedback

Rubric

feedback on

graphs and units

Data

literacy

Feedback

errors/overreliance

DIKSHA

verification and

Assessmen

t redesign

Viva

questions and

reflection

prompts

Oral

reasoning

Bias/unfairness Trustworthy

governance

principles

Note: Tables present design-oriented mappings; they should be adapted to local curricula and

resource contexts.

Table 2. Risk–mitigation matrix

Risk Why it matters Classroom

mitigation

Institutional

mitigation

Inaccurate

explanations

Misconceptions;

wrong causal models

Verification with

DIKSHA/NCERT;

unit checks

Tool validation;

vetted repositories

Academic

integrity

Assessment

invalidity

Process evidence;

viva; in-class tasks

Clear policy;

integrity regulations

Data privacy Student data

exposure

No PII;

anonymize

Privacy-by-

design; procurement

controls

Equity/access Unfair advantage Group use;

offline alternatives

Infrastructure

planning; inclusion

136

Bias/language

gaps

Unequal support Bias detection;

multiple sources

Indigenous

models; safe &

trusted AI

Table 3. Prompt templates with verification

Purpose Prompt template Verification output

Misconception check Explain ____ for Grade

__; list 3 misconceptions

and corrections; state

uncertainty

Corrections cited to

DIKSHA/NCERT

CER scaffold Convert this explanation

into Claim–Evidence–

Reasoning; do not add data

Evidence linked to data

table/graph

Data interpretation Propose 3 interpretations

+ 2 confounds; suggest

extra data

Chosen interpretation

justified with units/error

Inquiry planning Suggest

variables/controls and safe

procedure; risk checklist

Teacher-approved

procedure + checklist

Viva prep Generate 10 viva

questions incl. why/what-if

Oral answers assessed

Addendum B. Classroom Implementation Examples (Indian Context)

The examples below are aligned with inquiry-oriented, competency-based science learning and

can be adapted for different boards and state contexts. They are designed to keep scientific

reasoning and evidence at the center while using GenAI only as a scaffold.

Example 1 (Grades 6–8): Local Biodiversity Mini‑Inquiry

Topic: Classification, adaptations, ecosystems

Learning objectives:

 Observe and record local biodiversity in/near the school.

 Classify organisms using observable features.

 Write a CER explanation for one adaptation.

Lesson flow (5E-aligned):

 Engage: Generate “I wonder…” questions from campus photos.

 Explore: Field notes (8–10 observations).

 Explain: CER paragraph + evidence from notes.

 Elaborate: Compare two micro-habitats.

 Evaluate: 2-minute viva.

137

GenAI role (scaffold): GenAI generates field-note template and question prompts; students do

not use GenAI to identify species.

Prompt template: “Create a Grade 7 field-note template for biodiversity with columns for

observation, evidence, habitat, and classification.”

Verification requirement: Students cite at least one DIKSHA/NCERT source for concept

definitions.

Assessment idea: Portfolio + viva (integrity-by-design).

Example 2 (Grade 8): Water Filtration and Mixtures

Topic: Separation techniques; mixtures; environmental science

Learning objectives:

 Design a safe filtration procedure.

 Record observations and compare before/after.

 Discuss limitations and improvements.

Lesson flow (5E-aligned):

 Engage: Local water sources discussion.

 Explore: Build filter; collect observations.

 Explain: Graph turbidity proxy + CER.

 Elaborate: Improve design; justify changes.

 Evaluate: Viva + peer review.

GenAI role (scaffold): GenAI used for variable/control brainstorming and safety checklist,

under teacher approval.

Prompt template: “Suggest controlled variables, outcomes, and a safety checklist for a school

filtration experiment.”

Verification requirement: Teacher-approved protocol attached; no unsafe chemical instructions

permitted.

Assessment idea: Rubric lab report + transparency appendix.

Example 3 (Grades 9–10): Air Quality Data Reasoning

Topic: PM2.5/AQI trends, graph literacy, confounds

Learning objectives:

 Interpret a dataset; plot graphs with units.

 Propose explanations and confounds.

138

 Justify with data and references.

Lesson flow (5E-aligned):

 Engage: Provide week-long AQI values.

 Explore: Compute summaries and plot.

 Explain: Hypotheses + confounds list.

 Elaborate: Compare two locations.

 Evaluate: Oral questioning.

GenAI role (scaffold): GenAI generates alternative hypotheses and data needs; students choose

and justify.

Prompt template: “Given this time series, propose explanations, confounds, and additional data

to strengthen claims.”

Verification requirement: Students cite one trusted source (textbook/DIKSHA) for

pollution/health concept.

Assessment idea: In-class data task + viva.

Example 4 (Grades 11–12): Equilibrium Error‑Checking

Topic: Le Chatelier’s principle, constraints-based critique

Learning objectives:

 Predict equilibrium shifts.

 Detect errors in explanations.

 Justify corrections using equations.

Lesson flow (5E-aligned):

 Engage: Equilibrium scenario prompt.

 Explore: Individual prediction then pair compare.

 Explain: Critique AI explanation for errors.

 Elaborate: Create common-error cards.

 Evaluate: Written correction + viva.

GenAI role (scaffold): GenAI provides mixed-quality explanations that students audit.

Prompt template: Teacher: “Generate two explanations (one correct, one subtly incorrect) for

equilibrium shift; do not label.”

Verification requirement: Students cite definitions for Kc/Kp and justify corrections.

139

Assessment idea: Marks for error detection + corrected reasoning.

Example 5 (Undergraduate): Lab Report Integrity and Disclosure

Topic: Scientific writing, reproducibility, responsible tool use

Learning objectives:

 Submit raw data and first draft without GenAI.

 Use GenAI only for clarity and structure.

 Defend interpretations orally.

Lesson flow (5E-aligned):

 Draft 1: No GenAI.

 Revision: GenAI for language only.

 Append prompts/outputs and revision rationale.

 Viva to verify understanding and provenance.

GenAI role (scaffold): GenAI used for feedback generation; instructor validates before release.

Prompt template: “Provide rubric-aligned feedback on clarity; do not invent data; flag

uncertainties.”

Verification requirement: GenAI use statement required; student accountable for accuracy.

Assessment idea: Portfolio + viva; transparency graded.

Addendum C. Search Strategy: Databases, Search Strings, and Screening Workflow

This addendum provides a ready-to-use search strategy to operationalize the PRISMA 2020–

aligned approach described in the chapter (Page et al., 2021).

C1. Databases and Information Sources

Recommended academic databases (peer-reviewed literature):

 Scopus

 Web of Science Core Collection

 ERIC

 Google Scholar (supplementary)

 IEEE Xplore / ACM Digital Library (optional)

Recommended policy/grey literature sources (contextual grounding):

 UNESCO guidance on GenAI in education (UNESCO, 2023).

 OECD Digital Education Outlook sections on generative AI governance (OECD, 2023).

140

 Indian policy/curriculum: NEP 2020; NCF-SE 2023; NDEAR ecosystem policy;

DIKSHA resources.

 CBSE AI curriculum and integration manuals (CBSE, 2020; CBSE, 2024a).

C2. Example Boolean Search Strings (copy/paste ready)

Use publication years 2022–present for GenAI/LLM-focused searches, adjusting as needed.

Apply language limits only if justified.

Search focus Example Boolean string

Core GenAI + science education ("generative AI" OR GenAI OR "large

language model" OR LLM OR ChatGPT)

AND ("science education" OR biology OR

chemistry OR physics OR STEM) AND

(teaching OR learning OR assessment OR

inquiry OR laboratory)

Science writing / lab reports (ChatGPT OR "large language model"

OR LLM OR "generative AI") AND ("lab

report" OR "scientific writing" OR "science

communication") AND (students OR

classroom OR course)

Inquiry / experimentation ("generative AI" OR LLM OR

ChatGPT) AND (inquiry OR "inquiry-

based" OR experiment* OR "project-

based") AND (science OR STEM)

Assessment and integrity ("generative AI" OR ChatGPT OR

LLM) AND (assessment OR exam* OR

"academic integrity" OR plagiarism OR

cheating) AND (science OR STEM OR

education)

Teacher professional development ("generative AI" OR ChatGPT OR

LLM) AND (teacher* OR faculty) AND

("professional development" OR training

OR pedagogy) AND (science OR STEM)

Indian context filter (optional) ("generative AI" OR ChatGPT OR

LLM) AND ("science education" OR

STEM) AND (India OR Indian OR CBSE

OR DIKSHA OR NDEAR OR NEP)

Search string notes:

 Use truncation where supported (e.g., experiment*).

 In Scopus/WoS, consider field limits (TITLE-ABS-KEY or TS=).

 Add discipline terms (e.g., “chemistry education”).

 For multimodal GenAI, add “multimodal”, “image generation”, “prompting”, or “prompt

literacy”.

141

C3. Screening Workflow and Data Extraction Fields

Workflow: export results → deduplicate → title/abstract screening → full-text screening →

code and synthesize → report using PRISMA 2020 flow diagram and checklist (Page et al., 2021).

Extraction category Example fields

Bibliographic Author(s), year, country, venue

Context School/HEI; grade level; discipline;

region

Intervention Tool type; prompts; duration;

supervision

Design Qual/quant/mixed; sample; comparison

Outcomes Understanding; reasoning; writing;

engagement; performance

Risks/ethics Accuracy; bias; privacy; integrity; equity

Implementation Training; policy; infrastructure;

verification

Key findings Results + limitations +

recommendations

Ethics, Governance, and Academic Integrity (India-focused)

Responsible GenAI use in science education requires layered safeguards: classroom rules,

institutional policies, and system-level governance. UNESCO’s global guidance highlights data

privacy protection, transparency, and human agency, noting that educational systems should

validate GenAI tools for ethical and pedagogical suitability (UNESCO, 2023). NDEAR’s guiding

principles (privacy and security by design; interoperability) provide an Indian digital-architecture

lens for how GenAI could be deployed through controlled, auditable services rather than

unmanaged public tools (Ministry of Education, 2022).

Academic integrity is a central concern because GenAI can generate novel text that may bypass

traditional similarity-based plagiarism detectors. Indian HEIs already operate within UGC

academic integrity regulations; however, GenAI introduces new forms of ‘unattributed assistance’

that are not always captured by plagiarism definitions (University Grants Commission, 2018). A

practical response is to shift from purely product-focused assessment to process-focused evidence:

requiring drafts, lab notebooks, data provenance, reflective decision logs, and oral defence. This

approach both deters misuse and strengthens scientific reasoning as an assessed outcome.

Privacy and child safety are especially important in school settings. As a baseline, teachers

should enforce a no-PII (personally identifiable information) rule in prompts, use anonymized

student work for demonstrations, and prefer institutionally managed accounts where possible.

Schools can align GenAI adoption with existing digital policies and use DIKSHA resources for

142

verification so that learning does not depend on the accuracy of a single model output (DIKSHA,

2026).

Equity is a governance issue as well as a pedagogical one. If GenAI is used for homework or

take-home projects without ensuring access and guidance, it may widen existing achievement gaps.

Equity-first design includes guided in-class use, offline alternatives, and grading criteria that reward

reasoning and evidence rather than polished language alone. Multilingual scaffolds should be

validated to ensure that translation does not introduce scientific errors or cultural distortions.

Limitations of the Current Evidence Base

The evidence base on GenAI in science education remains emergent. Many studies are short-

term, focus on perceptions rather than objective learning outcomes, and are concentrated in higher

education contexts. Measurement approaches vary widely (writing quality, self-reported usefulness,

engagement proxies), making cross-study comparison difficult. Additionally, rapid tool evolution

(new model versions and features) can reduce the generalizability of findings across time.

Consequently, this chapter emphasizes design principles and governance-aligned practices that are

robust across tools, rather than tool-specific claims.

India-specific empirical studies on GenAI in school science are still limited. Contextual

factors—language diversity, device availability, class size, and local curricular constraints—likely

moderate GenAI’s effects. Therefore, pilot implementations should include local evaluation and

iteration, consistent with NEP 2020’s emphasis on context-appropriate evaluation before scale-up

(Ministry of Education, 2020).

Future Research Agenda for India

Future research in India should prioritize discipline-specific and longitudinal evaluation. Key

questions include: (1) Does GenAI-supported instruction improve conceptual understanding and

reduce misconceptions over time? (2) How does GenAI affect inquiry skills, including hypothesis

quality, variable control, and interpretation of uncertainty? (3) What forms of assessment redesign

best preserve integrity while promoting deep reasoning? (4) How do multilingual supports

influence science learning across home languages, and what validation routines are required? (5)

What governance models (privacy-preserving deployment, audit logs, institutional policies) work

best within NDEAR-aligned digital ecosystems?

Methodologically, design-based research can be valuable because it iteratively refines GenAI-

supported learning designs in real classrooms while collecting evidence on mechanisms and

outcomes. Large-scale teacher professional development studies—potentially delivered via

143

DIKSHA—can examine how teacher AI literacy, prompt design competence, and classroom

norms influence student outcomes and integrity incidents.

Conclusion

GenAI is poised to influence science education by changing how explanations are generated,

how writing is supported, and how inquiry activities are scaffolded. The evidence synthesized here

suggests that benefits are real but conditional: GenAI supports learning when embedded in

pedagogical designs that require verification, preserve learner agency, and align assessment with

reasoning. Risks—accuracy, integrity, privacy, and equity—are also real and must be addressed

through layered safeguards. India’s policy landscape (NEP 2020, NCF-SE 2023), digital

infrastructure (DIKSHA, NDEAR), and national AI capacity-building (IndiaAI Mission) provide

strong foundations for responsible adoption. The S‑CIENTIFIC framework and classroom

examples offered in this chapter provide practical pathways for Indian science educators to harness

GenAI as a scaffold for scientific temper and inquiry—without reducing learning to AI-generated

output.

144

References

Bettayeb, A. M., & Abu Talib, M. (2024). Exploring the impact of ChatGPT: Conversational AI

in education: A systematic review. Frontiers in Education, 9, 1379796.

https://doi.org/10.3389/feduc.2024.1379796

Central Board of Secondary Education. (2020). Artificial intelligence integration across subjects

for CBSE curriculum (Manual). https://www.cbse.gov.in/cbsenew/list-of-

manuals/AI_Integration_Manual.pdf

Central Board of Secondary Education. (2024a). Artificial Intelligence (Subject Code 417) Class

X: Curriculum for session 2024–2025.

https://cbseacademic.nic.in/web_material/Curriculum25/sec/417-AI-X.pdf

DIKSHA. (2026). Digital Infrastructure for Knowledge Sharing (DIKSHA) platform.

https://diksha.gov.in/

Digital India (MeitY). (2026). DIKSHA initiative overview.

https://www.digitalindia.gov.in/initiative/diksha/

IndiaAI. (2026). IndiaAI Mission: Portal and pillars. https://indiaai.gov.in/

Ministry of Education, Government of India. (2020). National Education Policy 2020.

https://ncert.nic.in/pdf/nep/NEP_2020.pdf

Ministry of Education, Government of India. (2022). National Digital Education Architecture

(NDEAR) ecosystem policy (Version 11 Nov 2022).

https://www.ndear.gov.in/images/pdf/NDEAR-Ecosystem%20Policy-Version.pdf

Ministry of Education, Government of India. (2023). National Curriculum Framework for

School Education 2023 (NCF-SE). https://www.education.gov.in/en/national-

curriculum-framework-school-education-2023

OECD. (2023). Emerging governance of generative AI in education. In OECD Digital

Education Outlook 2023. https://www.oecd.org/en/publications/oecd-digital-

education-outlook-2023_c74f03de-en/full-report/emerging-governance-of-generative-ai-

in-education_3cbd6269.html

Page, M. J., McKenzie, J. E., Bossuyt, P. M., Boutron, I., Hoffmann, T. C., Mulrow, C. D.,

Shamseer, L., Tetzlaff, J. M., Akl, E. A., Brennan, S. E., Chou, R., Glanville, J.,

Grimshaw, J. M., Hróbjartsson, A., Lalu, M. M., Li, T., Loder, E. W., Mayo-Wilson, E.,

McDonald, S., … Moher, D. (2021). The PRISMA 2020 statement: An updated guideline

for reporting systematic reviews. BMJ, 372, n71. https://doi.org/10.1136/bmj.n71

Press Information Bureau, Government of India. (2024, March 7). Cabinet approves ambitious

IndiaAI Mission to strengthen the AI innovation ecosystem (Press release).

https://www.pib.gov.in/PressReleaseIframePage.aspx?PRID=2012357

UNESCO. (2023). Guidance for generative AI in education and research.

https://www.unesco.org/en/articles/guidance-generative-ai-education-and-research

University Grants Commission. (2018). UGC (Promotion of Academic Integrity and Prevention

of Plagiarism in Higher Educational Institutions) Regulations, 2018.

https://www.ugc.gov.in/regulations/UGC_Regulations_university

Wang, X., Zainuddin, Z., & Leng, C. H. (2025). Generative artificial intelligence in pedagogical

practices: A systematic review of empirical studies (2022–2024). Cogent Education,

12(1), 2485499. https://doi.org/10.1080/2331186X.2025.2485499