LLM Foundation Course Legal Education and Research Methodology 2024-25 Question Paper with Solutions

Element two: the student bears responsibility for an outcome. This is what distinguishes a clinic from a demonstration. A student who watches a lawyer interview a client learns something; a student who conducts the interview, and whose omission will cost the applicant something, learns differently, because the consequences of a mistake attach to the learner.

Element three: supervision. Responsibility without supervision is not education, it is exposure of the client to an untrained person. The National Legal Services Authority (Legal Aid Clinics) Regulations, 2011 capture the concept exactly by requiring that a law college clinic operate under the supervision of a lawyer.

Element four: reflection. Doing without reflection produces habit, not understanding. The reflective conversation after the work is what converts an experience into a transferable competence, and it is the element most often missing from Indian internships, which is why an internship diary signed without discussion teaches nothing.

The models of clinic are conceptually distinct and should be distinguished. The live client clinic, in which students act for real applicants, is the strongest form, since all four elements are present. The simulation clinic, comprising moots, mock trials, client counselling and negotiation exercises, has the first, second and fourth elements and substitutes a constructed problem for a real one, gaining control and losing consequence. The externship or internship places the student in a practising office, and is clinical only if supervision and reflection are built in. The community legal literacy programme, the commonest Indian form, teaches by requiring the student to explain the law to people who need it, which tests understanding in the most demanding way available.

The Indian instantiation of the concept can then be stated briefly. The idea entered Indian legal education through the legal aid committees, that is the Krishna Iyer Committee (1973) on Processual Justice to the People and the Bhagwati Committee (1977) on National Juridical Care, rather than through the universities; it was demonstrated at scale by the National Law School of India University, established at Bangalore in 1987 with Prof. N.R. Madhava Menon as founding Director; it became compulsory when the Bar Council introduced the four practical papers in 1997; and it acquired a statutory place in the legal aid system through the NALSA regulations of 2011. Its urgency has increased since Bar Council of India v. Bonnie Foi Law College, 2023 SCC OnLine SC 130, in which a Constitution Bench overruled V. Sudeer v. Bar Council of India, (1999) 3 SCC 176, and upheld the All India Bar Examination, so that professional readiness is now tested by somebody other than the university.

State of Maharashtra v. Manubhai Pragaji Vashi, (1995) 5 SCC 730, tests the justice purpose of the concept. Maharashtra extended grant-in-aid to recognised private colleges but excluded recognised private law colleges. The Supreme Court held the exclusion arbitrary under Article 14, and reasoned that the duty to provide free legal aid under Article 39A read with Article 21 cannot be performed without competent lawyers and therefore without properly equipped law colleges. The concept is thereby shown to have a constitutional and not merely a pedagogic warrant: the State's duty to educate lawyers is derived from its duty to provide legal aid.

Suk Das v. Union Territory of Arunachal Pradesh, (1986) 2 SCC 401, tests the supervision element. The accused was tried and convicted unrepresented and without being told free legal services were available; the conviction was set aside, and the Court held that aid must be offered even where it is not asked for, identifying legal illiteracy as the obstacle. A clinic whose literacy work reaches people who do not know they have a right is doing what that decision said the State must do; a clinic without supervision would be adding an untrained adviser to a population already ill served, which is why supervision is an element of the concept and not a safeguard bolted on to it.

The concept measured, now that measurement is possible

A conceptual analysis is strengthened by noticing that the concept's central claim has recently become testable. Clinical legal education asserts that supervised practice with reflection produces a competence that exposition does not. Until 2023 that claim rested on institutional impression. After Bar Council of India v. Bonnie Foi Law College, 2023 SCC OnLine SC 130, decided 10 February 2023, in which a Constitution Bench overruled V. Sudeer v. Bar Council of India, (1999) 3 SCC 176, and upheld the All India Bar Examination, professional readiness is assessed by an examiner with no institutional stake in the answer. A concept whose claim can be checked is in better condition than one whose claim cannot.

Two further conceptual points complete the analysis.

The concept is not exhausted by the live client. A simulation clinic has the problem, the responsibility and the reflection, and substitutes a constructed problem for a real one, gaining control over what is taught and losing consequence. That is a genuine trade and not a deficiency: a moot can be designed to raise the precise issue a syllabus requires, and a real file cannot. The honest conceptual statement is that the models sit on a spectrum of reality, and each buys something by giving something up.

Conclusion. Conceptually, clinical legal education is the application of experiential learning to law, and it consists of four inseparable elements: a problem rather than a proposition, student responsibility for an outcome, supervision, and reflection. Its models differ in how much reality they admit, and it serves an educational purpose and a justice purpose which usually reinforce and occasionally conflict. India has adopted the concept fully in its regulations, through the four practical papers of 1997 and the legal aid clinic regulations of 2011, and partially in its classrooms, because the element most easily dropped, supervised reflection, is also the most expensive. A programme that keeps the syllabus and drops that element has kept the name of the concept and nothing else.

The method of doctrinal research is the identification of the relevant primary sources, their arrangement into a scheme, the extraction of the ratio decidendi from the decisions, the reconciliation or distinguishing of conflicting authority, the application of the rules of interpretation, and evaluation of the resulting statement against constitutional and jurisprudential standards. It is a real method with a real discipline, and describing it as "library work" understates it.

The method of non-doctrinal research is the definition of a universe, the design of a sample, the construction and pilot testing of an instrument, the collection of data by questionnaire, schedule, interview, observation, case study or record analysis, and its editing, coding, classification, tabulation and statistical analysis.

Merits of doctrinal research. It is the only method capable of stating the law authoritatively. It is quick and cheap, so it can be done well without funding, which matters in a system where most law faculties have none. It is replicable, because another competent reader can follow the same sources. It produces the systematic account on which teaching, practice and reform all rest. And it is indispensable when the law changes wholesale, as it did when the Bharatiya Nyaya Sanhita, 2023, the Bharatiya Nagarik Suraksha Sanhita, 2023 and the Bharatiya Sakshya Adhiniyam, 2023 came into force on 1 July 2024.

Demerits of non-doctrinal research. It is slow and expensive. It demands training in sampling, instrument design and statistics that the law curriculum does not provide. Access to court records, prisons and police stations depends on permissions. Official data is uneven, though the National Judicial Data Grid and the National Crime Records Bureau have improved matters. Findings are local. Respondents give socially acceptable answers. It carries ethical obligations and, since the Digital Personal Data Protection Act, 2023, statutory ones where personal data is collected. And a study whose doctrinal premise is wrong will measure the wrong thing with great precision.

The incentives that produced the imbalance, and the cost of it

A critical analysis has to explain why the Indian balance between the two methods is what it is, and the explanation is incentives rather than conviction.

Doctrinal work needs no funding, no permissions and no statistical training. It fits inside the time allowed for a dissertation, it can be completed by one person in a library, and it is easier to place in a journal. Empirical work fails on every one of those counts: it needs money for investigators, travel and translation; it needs access to court records, prisons and police stations, which is given inconsistently; it needs training in sampling and statistics that the LL.B. does not supply; and it takes longer than a dissertation allows.

That is why Upendra Baxi's criticism of 1975 has outlived every response to it. Any proposal to rebalance Indian legal scholarship that does not address funding, access to institutional data and methods training is an exhortation rather than a plan.

The court that could not find the research

The clearest demonstration of what the Indian imbalance costs is an occasion on which the Supreme Court needed empirical work and could not obtain it.

In Imtiyaz Ahmad v. State of Uttar Pradesh, (2012) 2 SCC 688, the Court was concerned with criminal proceedings stayed by High Courts and then left pending. Its examination of the figures showed that around nine per cent of such matters had been pending more than twenty years since the stay, roughly twenty one per cent more than ten, and that average pendency was about 7.4 years. It then asked whether delay was related to the strength of the judiciary. No reliable study existed, so the Court requested the Law Commission of India to examine the question, and the Commission produced its 245th Report, "Arrears and Backlog: Creating Additional Judicial (wo)manpower" (2014), which compared methods of calculating required judge strength and recommended a rate of disposal method in place of the judge to population ratio. A three judge Bench in 2017 directed interim computation of judge strength for each State pending the report of the National Court Management Systems Committee.

Three critical points follow.

The order the two are used in

The doctrinal chapter comes first, for a practical and not a hierarchical reason. An empirical study whose doctrinal premise is wrong measures the wrong thing with precision, and the error never shows up in the data. The transition of 1 July 2024, when the three criminal Sanhitas replaced the Penal Code, the Criminal Procedure Code and the Evidence Act, is the current illustration: no empirical question about the new law could even define its variables until the doctrinal question of what was carried over and what was renumbered had been answered.

Conclusion. Doctrinal research is authoritative, cheap and replicable, and it is structurally incapable of saying whether the law works. Non-doctrinal research answers exactly that question and is structurally incapable of stating the law. Their merits and demerits are therefore not competing claims but a description of a division of labour. The critical judgment to offer is about India specifically: in a legal system whose statutes are chiefly instruments of intended social change and whose characteristic failure is in implementation, the shortage of empirical work is not a stylistic preference of the academy but a gap in the knowledge available to legislators and courts, and it persists because the incentives inside the academy have never been altered to close it.

The questionnaire is a set of written questions completed by the respondent unaided. Procedure: decide the information needed from the hypothesis, draft short and unambiguous questions that are neither leading nor double barrelled, sequence them from easy to sensitive, add a preamble stating the purpose and the confidentiality undertaking, pilot on a small group and revise. Merits: cheapest per respondent, wide geographical reach, uniform wording, no interviewer bias, anonymity on sensitive matters, and time for the respondent to consult records. Limits: low and biased response rates, no control over who actually completes it, no clarification and no probing, unusable with illiterate respondents, and incomplete returns.

The schedule is the same instrument filled in by a trained investigator in the respondent's presence. Merits: far higher completion rates, usable where literacy cannot be assumed, and the investigator can clarify. Limits: cost, the need to train investigators, and the reintroduction of interviewer bias. In Indian conditions the schedule is the practical substitute for the questionnaire wherever the respondents are litigants rather than professionals.

The case study is the intensive examination of a single unit, whether one court, one scheme, one dispute or one community, using every available source. Merits: depth, the revealing of process and mechanism, suitability where the phenomenon is rare or not yet understood, the generation of hypotheses, and concrete detail that persuades. Limits: one unit cannot support a generalisation, selection bias, dependence on a single researcher, and slowness.

Documentary and record analysis is the systematic reading of court records, orders, charge sheets, institutional files, legislative debates and official reports against a coding frame. Merits: unobtrusive, cheap, capable of reaching the past, and grounded in what the institution itself recorded. Limits: records were created for another purpose, are frequently incomplete, and reflect recording practice as much as events.

Secondary statistical sources, that is the National Judicial Data Grid, the eCourts services, the National Crime Records Bureau, the census and the annual reports of the legal services authorities. Merits: large, cheap, already collected, and often the only route to a national picture. Limits: definitions are the collector's, not the researcher's; they change over time and between States; and they answer the question the collector was asking.

The obligations that now attach. Ethically, participation must be voluntary and informed, respondents must be told the purpose and how the data will be used, confidentiality must be maintained, and vulnerable respondents such as prisoners and children require particular care. Legally, where the data collected includes personal data, the Digital Personal Data Protection Act, 2023 governs its collection, storage and processing, so consent for a stated purpose, limitation to that purpose and secure handling are statutory requirements. Access to court records and to institutions requires permission, and the permission should be obtained in writing before the design is finalised.

Reliability, validity, triangulation and the published Indian sources

A "in detail" question is marked on the parts that a list of techniques omits, and there are three.

Reliability is consistency. An instrument or a coding frame that two people apply differently is unreliable, however sensible it looks, and it is tested by having two coders code the same returns and measuring their agreement.

One legal obligation now attaches to collection itself. Where the data includes personal data, the Digital Personal Data Protection Act, 2023, read against Justice K.S. Puttaswamy v. Union of India, (2017) 10 SCC 1, in which a nine judge Bench held informational privacy to be part of Article 21, makes consent for a stated purpose, limitation to that purpose and secure handling statutory requirements. Consent must therefore be built into the instrument and the storage plan into the design, not added at the writing stage.

Reliability, validity and the published Indian sources

A "in detail" answer on data collection is finished by the three concepts that decide whether the data is worth anything.

Reliability is consistency. A coding frame two research assistants apply differently is unreliable however sensible it appears, and it is tested by having two coders code the same returns and measuring their agreement.

Validity is whether the thing measured is the thing claimed. A court file records the date of disposal with perfect reliability and is an invalid measure of access to justice, because it is silent about disputes that never became files. The standing threat in legal research is the proxy accepted without argument.

One legal obligation now attaches to collection itself. Where the data includes personal data, the Digital Personal Data Protection Act, 2023, read against Justice K.S. Puttaswamy v. Union of India, (2017) 10 SCC 1, makes consent for a stated purpose, purpose limitation and secure storage statutory requirements, so consent belongs in the instrument and storage in the design rather than in an ethics paragraph at the end.

The six facts a report must disclose

A data collection chapter is checkable only if it states the universe, the sampling frame, the technique, the sample size, the response rate and the treatment of non-response. Without those the reader cannot tell which population has been described, and percentages become decoration. The same disclosure is what makes an honest non-probability design respectable: a study that states openly that it purposively selected ten advocates, and why, is usable, while one that reports the same ten as if they represented a district is not.

For full marks, cover: the vocabulary; why sampling is used; the probability techniques with a legal example and an analysis of when each is the right choice; the non probability techniques on the same plan; sample size; the two kinds of error; and a conclusion that gives a rule for choosing.

Sampling is the selection of a part of a population in order to draw conclusions about the whole. The population or universe is the entire set the researcher wishes to describe; the sampling unit is the individual element; the sampling frame is the list from which the sample is drawn; and the sample is the part actually studied. Getting these four terms right is the first thing an examiner looks for.

Sampling is used because complete enumeration is usually impossible and rarely necessary. A properly drawn sample of moderate size describes the population within a calculable margin of error, at a fraction of the cost, and often more accurately than a census, because fewer investigators mean better supervision of each one.

Stratified sampling. The population is divided into strata that matter to the question and a random sample is drawn within each. Example: separating represented from unrepresented applicants, or separating the courts of a district by seniority, before drawing. When to use it: where the strata differ from one another and are internally similar, and especially where a small stratum is important to the study and a simple random sample might miss it. Allocation may be proportionate, reflecting each stratum's share of the population, or disproportionate, deliberately over sampling the small important group and weighting the results afterwards.

Cluster and multi stage sampling. Naturally occurring groups are sampled, then units within them. Example: five districts at random, then three courts in each, then thirty files in each court. When to use it: where no frame of the whole population exists but a frame of the clusters does, and where fieldwork must be concentrated geographically, which describes almost every State wide legal study. The cost: less precision than simple random sampling for the same sample size, because units within a cluster resemble one another.

The non probability techniques give an unknown chance of selection and cannot support statistical generalisation, and yet a great deal of legal research has to use them.

Snowball sampling. Each respondent identifies the next. Example: released undertrials, victims of trafficking, or litigants who abandoned a legal aid application, none of whom appear on any list. When to use it: where the population is hidden or unlisted, which is common in exactly the research most worth doing. The weakness: the sample follows social networks, so an isolated member of the population is never reached.

Sample size is determined by four things and never by a fixed percentage: the variability of the population, the precision required, the confidence level chosen and the resources available. Where the analysis will break the sample into sub groups, the size of the smallest sub group usually decides the total, because a finding about a group of nine respondents is not a finding.

Size is settled by four things and never by a percentage of the population: the variability of the population, the precision required, the confidence level chosen, and the resources available. A fifth usually decides in practice, namely that every sub-group the analysis will report on must itself be large enough to support a conclusion.

The two errors must be distinguished, and this is the analytical core of the subject. Sampling error is the difference between the sample result and the true value arising because only a part was studied; it falls as the sample grows and can be estimated. Non-sampling error arises from a defective frame, non-response, bad question wording, investigator bias or coding mistakes; it does not fall as the sample grows and cannot be estimated.

What must always be disclosed is the universe, the frame, the method, the size, the response rate and the treatment of non-response. Percentages without those six facts are decoration.

A worked design, and the disclosure that makes it checkable

The techniques are worth little until one design is carried through, because the frame and not the researcher's preference decides the method.

The question: do aided litigants obtain decisions on merits at the same rate as unaided ones? The universe is all matters instituted in a district's civil courts in a stated year. The frame is the institution register, which exists and is complete, so a probability method is open. The method is stratified sampling, because representation status is the variable of interest and a simple random sample might return too few aided matters to analyse; the register is split into aided and unaided strata and a random sample drawn within each, with disproportionate allocation over-sampling the smaller stratum and weighting the results afterwards. The size is fixed by the precision required and by the rule that the smallest sub-group reported on must itself support a conclusion.

Conclusion. The sampling techniques available divide into probability methods, that is simple random, systematic, stratified and cluster or multi stage sampling, which permit generalisation with a stated margin of error, and non probability methods, that is purposive, quota, convenience and snowball sampling, which establish possibility rather than frequency. The rule for choosing is straightforward: if the study must estimate how common something is, a probability method and an honest frame are indispensable; if it must understand how something works, a purposive selection is not a compromise but the correct design. What is never acceptable is to use a non probability sample and then report the result as though it described a population.

• To prevent duplication. Discovering after a year that the question was answered in 2019 is the most expensive avoidable error in postgraduate work.
• To identify the gap that the study will occupy, which is what converts a subject into a research problem.
• To supply the theoretical framework, without which a finding cannot be connected to anything already known.
• To reveal the variables that other researchers have found to matter, which is what allows a hypothesis to be more than a guess.
• To show the methods used on comparable problems, and which of them failed.
• To provide the standard of comparison against which the study's own findings will be reported.
• To establish the researcher's competence in the field, which is what a supervisor and an examiner are assessing when they read the chapter.

Its limits. It is never complete, and a researcher can read indefinitely instead of beginning. Access to Indian material is unequal because the commercial databases are expensive, though eSCR, India Code and open access repositories have narrowed the gap. And immersion in a literature carries the risk of adopting its assumptions along with its findings.

b. Formulation of hypothesis

A hypothesis is a tentative proposition about the relation between two or more variables, framed before the investigation and in a form that evidence can defeat. Its two essential features are that it is provisional and that it is falsifiable, and both follow from the fact that its purpose is to be tested rather than defended.

Formulation runs through six steps.

1. Narrow the problem until it names a population, a place and a period.
2. Review the literature, which is the previous half of this question and is what supplies the candidate variables.
3. Identify the variables, dependent and independent.

The characteristics of a workable hypothesis are conceptual clarity, empirical testability, specificity as to population, place and period, feasibility with the data and techniques available including lawful access under the Digital Personal Data Protection Act, 2023, connection to a body of theory, simplicity, consistency with established facts, and neutrality of wording.

Three errors to avoid. Stating a conclusion instead of a hypothesis, that is asserting what the researcher already believes and designing the study to confirm it. Stating a proposition no observation could contradict. And revising the hypothesis after the data has been seen so that it fits the result, which is fitting and not testing.

The exception. Doctrinal and genuinely exploratory research proceed on a clearly stated research question rather than a hypothesis, and the report should say so rather than manufacture one to fill a heading.

The two limbs joined, and the errors each invites

The paper sets these two together because they are consecutive stages, and the answer is stronger for saying so and then showing where each goes wrong.

The operational definition is the joint between the two limbs and is where most marks are lost. The literature supplies the candidate variables; the hypothesis is testable only once those variables are defined in terms of what will actually be counted. "Effectiveness of legal aid" is a phrase from the literature; "the proportion of aided matters disposed of on merits within eighteen months" is a variable, and only the second can appear in a hypothesis.

Two errors of testing complete the answer. A Type I error is the rejection of a true null hypothesis, that is finding a difference that does not exist, and the significance level fixed in advance is the probability of it the researcher accepts. A Type II error is the failure to reject a false null, that is missing a difference that does exist, and it is controlled chiefly by sample size. Because reducing one increases the other, the choice of level is a judgment about which mistake would matter more, and in research bearing on policy it should be stated rather than assumed.

Where the hypothesis comes from, and the check that ends formulation

Both limbs are finished by the same point: a proposition has to come from somewhere, and it has to be capable of being abandoned.

The literature review is what makes the source usable. Reading tells the researcher which variables others have found to matter, so that a hypothesis is a considered proposition rather than a guess, and it identifies the gap into which the proposition must fit. A review that lists what each writer said, without recording where they disagree and what none of them answered, has done the reading and not the review.

The check that ends formulation is a single question, and it disposes of most bad hypotheses. Ask what finding would make the researcher abandon the proposition. If nothing comes readily to mind, it is not falsifiable and the study will produce a defence of a belief rather than a test. The corollary is that the hypothesis must be recorded before collection begins, in the research proposal and in the supervisor's file, because a hypothesis rewritten after the data has been seen so that it fits the result has been fitted and not tested, and the difference is invisible in the finished report unless the earlier version exists.

For full marks, cover: what ICT covers here; its role at each stage of research, which is the clearest organising device; the Indian sources by name, because that is what makes the answer specific; the different effect on doctrinal and empirical work; the risks, including artificial intelligence and data protection; and a conclusion on what it has not changed.

Information and communication technology in legal research means the electronic tools by which legal material is located, verified, collected, analysed, cited and communicated. Its role is best set out stage by stage, because its contribution is uneven across the research process.

Locating the problem and reviewing the literature. Full text search does not depend on an indexer having anticipated the question, so a phrase appearing anywhere in a judgment can be found, including in argument, in a dissent or in an obiter passage no digest would carry. Bibliographic databases, open access repositories and institutional archives make a literature review possible from anywhere, and the Annual Survey of Indian Law, the Law Commission reports and committee reports are all online.

Collecting empirical data. Online survey platforms reach a dispersed sample at negligible cost. Institutional data that once required a visit is downloadable, and the National Judicial Data Grid in particular has made a whole class of Indian studies on delay and pendency possible for a researcher with no funding at all.

Analysing. Statistical packages such as SPSS and R, and spreadsheets for smaller studies, perform in minutes the cross tabulation and significance testing that once made empirical legal work impractical for one person. Qualitative software assists with the coding of interview transcripts.

Writing and citing. Reference managers such as Zotero and Mendeley maintain a citation database and generate footnotes in a chosen style, removing the commonest source of error in a dissertation. Similarity detection software is standard, because the UGC (Promotion of Academic Integrity and Prevention of Plagiarism in Higher Educational Institutions) Regulations, 2018 require institutions to run such checks.

Communicating. Open access journals and repositories have widened the readership of Indian legal scholarship far beyond the libraries that subscribe to printed journals.

Incomplete coverage. Databases differ, particularly on older High Court decisions and on subordinate legislation, so a negative result in one is not proof that nothing exists.

Authenticity. The online text is a finding aid. The authoritative version of a statute is the Gazette text and of a judgment the official report, and where a point turns on the exact words the authoritative version governs.

Link decay. Web pages change and disappear, which is why a webography entry must carry the full address and the date of access.

Artificial intelligence. Generative tools draft fluently and can cite decisions that do not exist. They may be used to summarise and to draft, and every authority they produce must be verified against the report before it is used. The obligation is the researcher's and cannot be delegated to the tool.

Data protection. Where research collects personal data electronically, the Digital Personal Data Protection Act, 2023 governs consent, purpose limitation and secure storage.

Artificial intelligence raises the stakes rather than lowering them. Generative tools produce fluent, confident text and will sometimes cite decisions that do not exist. They may properly be used to summarise, to draft and to translate, and every citation they produce must be checked against the report. The duty belongs to the researcher, and a dissertation citing a fabricated authority is dishonest on its face rather than merely mistaken.

What ICT has not changed is the right note to end on. It has not supplied judgment: choosing the question, reading the authority in full rather than the headnote, noticing that it has been overruled and quoting the provision in the words the Gazette uses are unchanged obligations. It has not equalised access completely, since subscription databases remain expensive even though eSCR, India Code and the National Judicial Data Grid have narrowed the distance between a national law university and an ordinary college. And it has not removed the bias in the frame: a database of reported judgments, however searchable, is still a record of a small and unrepresentative fraction of what courts do, so an ICT-enabled study of "judicial trends" inherits exactly the sampling problem it appears to solve.

The problem ICT does not solve, and can conceal, deserves the same emphasis. A searchable database of reported judgments remains a record of a small and systematically unrepresentative fraction of what courts do, weighted towards appeals, represented parties and contested matters. A study of judicial trends conducted entirely inside such a database inherits the frame defect it appears to have escaped, and the ease of retrieval makes the defect harder to notice, because the researcher now holds hundreds of decisions where a digest would have yielded a dozen. The corrective is the institutional data above, which describes the system rather than its reported output.

Two duties survive every improvement in the tools. The citator must be run, because a decision may have been overruled, as V. Sudeer v. Bar Council of India, (1999) 3 SCC 176, was by Bar Council of India v. Bonnie Foi Law College, 2023 SCC OnLine SC 130. And every citation produced by a generative artificial intelligence tool must be opened and read in the report before it is used, because such tools generate authorities that do not exist and do so in a format indistinguishable from a real one.

Conclusion. Information and communication technology has changed the economics of legal research: retrieval that once depended on the indexer is now direct, currency that was once assumed is now verifiable, comparative and empirical work that once required an institution is now within reach of an individual, and free Indian portals such as eSCR, India Code and the National Judicial Data Grid have narrowed the gap between a well funded institution and a poor one. What it has not changed is the researcher's obligation to choose the right question, to read the authority in full and to confirm that it still stands. The easier collection has become, the more the discipline of verification carries the weight of the work.

Its types. A structured interview puts fixed questions in a fixed order and is closest to a schedule. An unstructured interview is guided only by a topic. A semi structured interview follows a list of themes and allows the interviewer to probe, and it is the workhorse of socio-legal research because it combines comparability with depth. A focused interview concentrates on the respondent's experience of a particular event, such as a hearing. A group interview or focus group puts the question to several respondents together, which produces interaction but suppresses dissent.

Its conduct. Prepare a guide of themes drawn from the hypothesis; secure informed consent and explain how the material will be used; begin with easy questions and reach sensitive ones late; probe a thin answer rather than accept it; avoid leading; record with permission and transcribe promptly, because memory of tone and emphasis fades within a day.

Its merits. Depth, and the ability to follow an unexpected answer. Clarification of a misunderstood question. Access to material no form could elicit, such as why a complainant turned hostile. Non verbal information and the setting. High response rates. And usability with respondents who cannot read, which in Indian legal research is decisive.

Its types. A census survey covers the whole population; a sample survey covers a part. A cross sectional survey takes a snapshot at one moment; a longitudinal survey returns to the same population over time, and a panel study returns to the same respondents. A descriptive survey establishes what is the case; an analytical survey seeks to explain it.

Its use in legal research. It is the natural design for questions about the incidence of a practice, the awareness of a right, the experience of litigants and the opinions of professionals. Indian examples include surveys of awareness of the free legal services entitlement under section 12 of the Legal Services Authorities Act, 1987, of the working of internal assessment in law colleges, and of the referral of matters to mediation.

Its merits. It can describe a large population from a modest sample; it produces countable data; it permits comparison between groups and over time; and it is the only method that can estimate how common something is.

The steps. Select a broad field the researcher can live with for the length of the work. Read widely in it, because a problem is recognised and not invented. Look at the recognised sources of legal problems: a conflict between High Courts, a dissenting judgment, a provision amended in response to judicial construction, a Law Commission recommendation not acted upon, a newly enacted statute whose working is untested, and the distance between what a statute promises and what practitioners report. Discuss the candidate problem with the supervisor and with people in the field. And survey the available material to confirm that the sources exist and can be reached.

The criteria. Novelty, significance, feasibility in time and access, clarity, ethical acceptability including compliance with the Digital Personal Data Protection Act, 2023 where personal data is collected, and a scope neither so wide that it produces description nor so narrow that it produces nothing.

Why it matters. The problem fixes the objectives, the hypothesis, the design, the sources and the method, so a badly stated problem cannot be repaired by good work later. A problem too wide produces description; one already answered produces a literature review presented as research; one whose data cannot be obtained produces a dissertation that silently changes its subject.

The Indian material to know by name includes the Journal of the Indian Law Institute, the Annual Survey of Indian Law, the National Law School of India Review, the Indian Journal of International Law, and the reports of the Law Commission of India, which are juristic writing with an official character.

Its limits. It dates from the day it is printed. It carries the author's standpoint. And it is not authority, so every proposition taken from a commentary must be traced back to the statute or the judgment before it is relied on. The working rule is to use juristic writing to find and evaluate authority, and never to cite it in place of the authority itself.

What each note is actually tested on

Four notes, two to be written, and each has a distinguishing point that separates a strong answer from a general one.

Identification of the research problem is tested on the distinction from a topic and on the Indian sources of problems: a conflict between High Courts, a dissenting judgment, a provision amended in response to judicial construction, an unimplemented Law Commission recommendation, and an untested statute, of which the three criminal Sanhitas in force from 1 July 2024 and the Mediation Act, 2023 are the current examples.

Juristic writing is tested on use and limit. Its uses are to enter a field quickly, to supply criticism, to identify the gap, to supply the theoretical framework and to name new concepts before the courts adopt them; its persuasive weight in India is real, since courts cite Mulla, Seervai, Basu, and Ratanlal and Dhirajlal. Its limit is that it is not authority: every proposition must be traced back to the statute or the judgment, and it dates from the day it is printed, which is why the Annual Survey of Indian Law rather than a treatise is the instrument for finding where a subject has recently moved.

The four are one sequence in disguise, and saying so is worth a mark: juristic writing tells the researcher what has already been concluded, identification of the problem states what has not, and the survey and the interview are the two instruments that answer it, the first measuring how widespread something is and the second explaining why.

The discipline that runs through all of them is disclosure. An interview based finding is worth what the reader is told about how the respondents were chosen and what was asked. A survey finding is worth what the reader is told about the universe, the frame, the sampling method, the size and the response rate. A research problem is worth what the reader is told about the literature that established it was open. And a proposition drawn from juristic writing is worth nothing at all until it is traced to the statute or the judgment behind it, because a commentary is dated from the day it is printed, and the replacement of the criminal codes on 1 July 2024 turned a shelf of standard works into legal history.

One caution belongs to two of the four and is now statutory. Interviews and surveys collect personal data, and since the Digital Personal Data Protection Act, 2023, read against Justice K.S. Puttaswamy v. Union of India, (2017) 10 SCC 1, in which a nine judge Bench held informational privacy to be part of Article 21, consent for a stated purpose, purpose limitation and secure storage are legal requirements rather than courtesies. That belongs in the design and in the preamble to the instrument, not in a paragraph of acknowledgements at the end.