By Manal Helal on 29/5/2026

Abstract

The phrase “that is not computer science” is often used to police disciplinary boundaries. This article uses a student’s description of some teaching as “non-orthodox computer science” to examine a different problem: the unadapted importation of methods from other disciplines into computer science education. The article argues that interdisciplinary teaching can enrich computer science when aligned with the discipline’s core practices, but can weaken student development when it displaces implementation, debugging, empirical evaluation, benchmarking, and iterative improvement.

The article also distinguishes constructive interdisciplinarity from ethically problematic borrowing. It discusses examples in which game-theoretic or reinforcement-learning models can encode unequal payoffs for social groups. Such work requires careful ethical framing, human rights impact assessment, and caution before any translation into policy or deployment in systems.

The article compares research methods and pedagogical practices across social sciences, mathematics, engineering, medical sciences, basic sciences, and computer science. It then identifies shared educational principles—constructive alignment, authentic assessment, formative feedback, active learning, outcome-based education, and research–teaching integration—while arguing that computer science remains methodologically distinctive as an artefact-centred, hybrid design science with a strong culture of implementation and empirical comparison.

The article concludes that genuine interdisciplinarity depends on strong disciplines rather than their erasure. Defending the pedagogical identity of computer science is not an argument against collaboration; it is a condition for preparing graduates who can contribute computational expertise to other fields.

Keywords: computer science education, disciplinary identity, interdisciplinarity, antidisciplinarity, research methods, pedagogy, assessment, algorithmic discrimination, competency-based education, constructive alignment. 

Introduction

“That is not computer science.”

The phrase “that is not computer science” carries disciplinary force. As Matti Tedre recounts in the preface to The Science of Computing: Shaping a Discipline (2014) [1], it has been used to reject doctoral theses, tenure cases, and funding applications. Tedre’s response was not simply to defend a fixed boundary, but to ask what computer science is and why its identity has been difficult to define.

My starting point is related but inverted. The concern did not come from a traditional computer scientist defending canonical boundaries. It came from a student in a computer science department who described some teaching as “non-orthodox computer science.” The concern was not that colleagues from other fields lack value. Rather, it was that methods appropriate in mechanical engineering, social science, mathematics, or biology may not automatically serve computer science students unless they are adapted to computational aims and assessment practices.

For example, some students are encouraged to practise artificial intelligence algorithms mainly through pen-and-paper exercises, even when the intended competencies require implementation, empirical evaluation, and interpretation of model performance. Similarly, project feedback such as “not deep enough” or “can be better” is too general to guide technical improvement. In computer science, effective feedback should usually identify concrete issues in problem formulation, code quality, experimental design, benchmarking, reproducibility, or interpretation of results.

This matters because computer science graduates are expected to design, implement, test, debug, benchmark, and iterate on computational artefacts. The field changes quickly, particularly in areas such as machine learning and large-scale software systems. Students, therefore, need habits of rapid experimentation and evidence-based evaluation, not only conceptual understanding.

The Dark Side of Interdisciplinary Borrowing – When Algorithms Enable Oppression

Not all interdisciplinary exchanges are equally beneficial. Some uses of evolutionary game theory and reinforcement learning model social groups through unequal payoff structures. Such models may be valuable for understanding how discriminatory norms emerge. Still, they become ethically problematic when the modelling choices are presented as natural, desirable, or suitable for policy without explicit normative safeguards. One example is the paper by McElreath, Boyd, and Richerson (2003) [2], “Shared Norms and the Evolution of Ethnic Markers,” which models ethnic markers and behavioural strategies under particular payoff assumptions. The concern is not that modelling social behaviour is inherently illegitimate, but that models of group differentiation require careful explanation of assumptions, limits, and ethical implications. 

A related concern arises in reinforcement learning. RL agents optimise reward functions through interaction with an environment. If an environment rewards resource hoarding, exclusion, or unequal treatment of groups, an agent may learn policies that reproduce those incentives. This risk is especially important in domains such as resource allocation, recommendation, finance, security, or public-sector decision support. The appropriate conclusion is not that RL is inherently oppressive, but that reward design, constraint specification, auditing, and deployment governance are central ethical responsibilities. Where algorithmic systems are used to make decisions about people, discriminatory objectives or outcomes may conflict with human-rights and anti-discrimination principles, including non-discrimination and equal protection under the Universal Declaration of Human Rights [3] and relevant domestic or regional legislation (e.g., the US Civil Rights Act[4], the UK Equality Act[5], the EU Charter of Fundamental Rights[6]). A simulation or theoretical model is not inherently unlawful; however, its deployment or use in policies that disadvantage people because of protected characteristics can pose serious legal and ethical risks. 

When a mathematical model or an RL reward function deliberately reduces outcomes for people based on ethnic markers or other protected characteristics, it should be treated as a high-risk design choice. The central issue is whether the work describes discrimination for critical analysis or normalises discriminatory treatment as an optimisation objective. The broader lesson is that interdisciplinary borrowing requires accountability. Borrowing statistical, mathematical, or computational methods can strengthen research when assumptions are transparent and disciplinary standards are preserved. It becomes problematic when methods are used to legitimise unequal treatment or when ethical constraints are treated as external to the technical problem. Computer science departments should therefore teach students not only how to build algorithms, but also how to assess the social and legal implications of reward functions, datasets, benchmarks, and deployment contexts. 

Therefore, in the broader debate about interdisciplinarity, we must draw a clear line. It is one thing to borrow statistical methods from biology or physics to improve computer vision. It is quite another to borrow evolutionary game theory to normalise ethnic favouritism, or to deploy RL as a tool for suppressing emergent social movements. These are not legitimate interdisciplinary research directions; they are abuses of computational methods that should be actively monitored, called out, and – where they cross into policy or deployment – legally challenged. As I have argued in my analysis of the McElreath et al. paper [7], the failure to explicitly reject such models as unethical, or to demand that human rights impact assessments accompany them, is itself a failure of academic responsibility. Computer science departments, in particular, have a duty to teach not only how to build algorithms, but also how to recognise and refuse research that weaponises our discipline against vulnerable groups.

Article Aims

This article, therefore, has three purposes. First, I revisit and extend the analysis of research and teaching methodologies across disciplines – social sciences, mathematics, engineering, medical sciences, basic sciences, and computer science – to make explicit why each field’s methods are well-suited to their own purposes but not automatically transferable. Second, I argue that computer science is not only a distinct discipline with its own research and teaching practices, but also a service discipline to all others. We contribute algorithms, infrastructure, and computational thinking to every field – from biology to linguistics to economics. However, being a service discipline does not mean dissolving into others. To remain a strong partner, we must maintain our own competitive graduate attributes: the ability to code, evaluate empirically, deploy systems, and adapt rapidly to change. Third, I situate this discussion within the broader higher education debate about interdisciplinary versus anti‑disciplinary trends. As scholars have warned, the uncritical push for interdisciplinarity can sometimes devolve into what Paul Griffiths (2022) calls “antidisciplinarity” – in which the dilution of disciplinary standards is mistaken for innovation and administrative convenience masquerades as intellectual progress [8]. The deliberate chaos that some celebrate actually undermines the very conditions that enable rigorous, replicable, and fast‑paced progress. As the history of higher education in the United Kingdom shows, genuine interdisciplinarity rests upon strong disciplines, not their erasure [9]. This concern is further studied by a large‑scale empirical study of nearly half a million university syllabi from 2004 to 2019 [10]. Using natural language processing, the study found that despite decades of institutional rhetoric, university teaching has remained stubbornly disciplinary: terminology, topic distributions, and even pedagogical verbs (e.g., “apply,” “understand”) showed no significant shift toward interdisciplinarity. The only exception was engineering and technology, which modestly incorporated social and natural science content. The authors conclude that interdisciplinary education is “still largely rhetorical,” and that universities suffer from a decoupling between their research ambitions and their everyday teaching practices. In other words, the deliberate chaos that some celebrate actually undermines the very conditions that enable rigorous, replicable, and fast‑paced progress – and this is nowhere more evident than in the mismatch between how computer science should be taught and how imported methods from other disciplines are diluting it. 

I write not as an enemy of interdisciplinarity, but as a defender of disciplinary integrity within it. My colleagues from mechanical engineering, social science, maths, and biology bring valuable perspectives. But their methods – however excellent for their own fields – should not become the straitjacket for computer science education. The student who complained about “non‑orthodox computer science” was not being narrow‑minded. They were recognising something fundamental: that computer science has its own way of asking questions, designing solutions, and evaluating outcomes. To suppress that recognition in the name of interdisciplinarity is not progress. It is a failure of academic leadership.

The challenge, then, is not to build walls between disciplines. The challenge is to understand their differences deeply enough to collaborate without erasing what makes each field effective. This article is an attempt to map those differences – in research methods, in pedagogical practices, in assessment philosophies – and to argue for a computer science education that is disciplinary‑confident: one that serves others not by becoming like them, but by being unapologetically itself.

This initial draft article provides a structured differentiation of research methodologies across disciplines, contrasted with computer science (CS). The analysis follows the arc from question conceptualisation → aims/objectives → method plan → examples of impactful papers → explicit differences with CS. Finally, CS’s own research process is detailed. This synthesis shows that while CS shares tools with other fields, its object of study (computational artefacts) and evaluation logic (performance over benchmarks, formal guarantees, or user studies) make it methodologically distinct – a design science with empirical and theoretical branches. Curriculum frameworks such as the ACM/IEEE-CS/AAAI Computer Science Curricula [11] are routinely updated because computer science must keep pace with its own evolution—especially since most impactful research now comes from industry. Continued alignment with this pace is more than sufficient; imposing new pedagogical constraints from other disciplines adds no value. Such constraints only disadvantage students, preventing them from keeping up with a field whose pedagogy shaped the very learning styles of computer science academics themselves. 

1. Social Sciences vs. Computer Science

Research Process

• Question conceptualisation: Rooted in ontological (nature of reality) and epistemological (nature of knowledge) positions – interpretivism vs. positivism. Questions often ask “why” and “how” about human behaviour, institutions, or societies [11].
• Aims & objectives: Typically exploratory, descriptive, or explanatory; often seek causal mechanisms or thick description [12].
• Methods: Qualitative (ethnography, interviews, grounded theory) or quantitative (surveys, quasi-experiments, longitudinal panels). Triangulation is common.
• Example impactful paper: Granovetter (1973) – “The Strength of Weak Ties.” Used qualitative interviews and network mapping of job seekers; changed sociology and network theory [13].
• Differences with CS: Social science prioritises contextualised interpretation, researcher reflexivity, and often small-N studies. CS prioritises reproducibility, scalability, and formal models. Social science rarely uses “benchmark datasets” as ground truth.

Social Sciences Critical Pedagogy and Authentic Assessment

Pedagogical Philosophy

Social sciences education is rooted in critical pedagogy and constructivism—the idea that knowledge is socially constructed and that students must engage with multiple, often conflicting, perspectives [14]. The goal is not merely content acquisition but the development of critical citizens who can question prevailing narratives and understand structural inequalities.

Best-practice pedagogies:

• Problem-based learning (PBL): Students grapple with real-world social problems (e.g., poverty, inequality) and must draw on multiple disciplinary lenses to propose solutions.
• Collaborative inquiry and debate: Structured classroom discourse that exposes students to diverse viewpoints and requires evidence-based argumentation.
• Reflexive practice: Students critically examine their own positionality and biases as knowers.
• Community-engaged learning: Partnerships with community organisations to apply theoretical knowledge to authentic social issues.

Assessment Practices

The shift away from standardised testing toward authentic assessment is pronounced in the social sciences [14], [15]. The concept of “authentic pedagogy” emphasises four dimensions:

1. Disciplinary depth – demonstrating substantive knowledge of the field’s concepts and methods.
2. Depth of analysis – going beyond description to explanation, critique, and synthesis.
3. Richness of communication – producing elaborated, well-structured written and oral products.
4. Recognition of knowledge’s problematic nature – acknowledging uncertainty, multiple interpretations, and the constructedness of claims [14], [15].

Typical authentic assessments:

• Research papers and literature reviews
• Policy briefs and case study analyses
• Reflective journals and position papers
• Oral presentations and debates
• Portfolio-based assessment across a programme of study

Undergraduate vs Postgraduate Differences

Level	Focus	Pedagogy	Assessment
Undergraduate	Foundational concepts, disciplinary socialisation, critical thinking	Lecture-discussion hybrids, structured PBL, guided inquiry	Essays, exams with short answer/essay, annotated bibliographies
Postgraduate	Original research, theoretical sophistication, methodological expertise	Research seminars, independent projects, and doctoral supervision	Thesis/dissertation, publishable papers, grant proposals, conference presentations

2. Mathematics vs Computer Science

Research Process

• Question conceptualisation: Starts with conjectures, open problems, or gaps in logical structures (e.g., number theory, topology). Questions are purely abstract: “Does there exist…” or “Under what conditions…”
• Aims & objectives: Prove or disprove a theorem, establish necessary/sufficient conditions, or classify objects up to isomorphism.
• Methods: Axiomatic reasoning, proof by contradiction, induction, construction, or counterexample. No empirical data. Peer review via formal proof checking.
• Example impactful paper: Gödel (1931) – “On Formally Undecidable Propositions…” – changed the foundations of mathematics and computability [16].
• Differences with CS: Mathematics is non-empirical; truth is eternal and deductive. CS, even in its theoretical form, often involves algorithmic complexity relative to computational models (e.g., Turing machines) and occasionally empirical validation (e.g., runtime experiments). CS theorems (e.g., NP-completeness) are applied to problems with input data.

Mathematics Pedagogical: Rigour, Proof, and Conceptual Understanding

Pedagogical Philosophy

Mathematics education at the undergraduate level emphasises conceptual understanding over rote procedural fluency. The MAA Instructional Practices Guide identifies three foundational practice types: classroom practices, assessment practices, and design practices, all of which are informed by empirical research [17].

Best-practice pedagogies:

• Inquiry-based learning (IBL): Students discover mathematical truths through guided investigation rather than passive lecture.
• Process-oriented guided inquiry learning (POGIL): Structured team-based activities where students construct their own understanding.
• Flipped classroom: Students engage with lecture content outside class, using class time for problem-solving and discussion.
• Collaborative problem-solving: Students work in small groups to solve challenging problems, articulating their reasoning and critiquing others’ approaches.

Assessment Practices

Mathematics assessment is distinctive in its emphasis on process validity—ensuring that assessments measure not just correct answers but mathematical reasoning, proof construction, and problem-solving strategies [18].

Key assessment types:

• Problem sets: Regular assignments requiring step-by-step solutions with justifications
• Examinations: Often combining computational problems with proof-based questions
• Oral examinations (vivas): Students verbally articulate their reasoning, especially at the postgraduate level
• Portfolios: Collections of polished proofs and extended problem solutions
• Peer instruction: Students assess and provide feedback on each other’s proofs

Formative Assessment

Frequent, low-stakes assessment with immediate feedback is critical in mathematics, where misconceptions can compound rapidly. The emphasis is on identifying where reasoning breaks down, not merely marking answers as correct or incorrect.

Undergraduate vs Postgraduate Differences

Level	Focus	Pedagogy	Assessment
Undergraduate	Core mathematical language, proof techniques, problem-solving	Lecture with active learning components, recitations, and problem sessions	Weekly problem sets, midterm/final exams, and possibly a senior project
Postgraduate	Advanced specialisation, original theorem proving, research methods	Research seminars, reading courses, supervised thesis work	Qualifying exams, thesis, paper submissions, and teaching assistantship

3. Engineering Disciplines (Mechanical, Electrical, Civil) vs Computer Science

Research Process

• Question conceptualisation: Driven by real-world constraints: physical laws, material properties, safety factors, cost, energy efficiency—questions: “How to design X to withstand Y?” [19].
• Aims & objectives: Optimise a system, predict failure modes, or validate a prototype under specified conditions.
• Methods: Modelling & simulation (FEA, CFD), physical prototyping, wind/water tunnels, destructive testing, empirical measurement. Often iterative: design–build–test–redesign.
• Example impactful paper: Wright brothers’ wind tunnel data (1902–1903) – empirical airfoil tests that led to the first powered flight.
• Differences with CS: Engineering disciplines require physical instantiation and handle continuous variables, noise, and material tolerances. CS works with discrete logic, software abstraction, and bit-exact reproducibility. CS prototypes can be duplicated at zero marginal cost; physical prototypes cannot.

Engineering Pedagogy: Outcome-Based Education and Competency Development

Pedagogical Philosophy

Engineering education is dominated by Outcome-Based Education (OBE), a paradigm shift from teacher-centric content delivery to student-centric skill acquisition [20]. All teaching and assessment is backwards-designed from clearly articulated learning outcomes, which are themselves aligned with program accreditation requirements (e.g., ABET in the US, Washington Accord internationally).

The SPLAM-OBE Framework integrates five pedagogical strategies for engineering education [20]:

1. Active-learning – students engage meaningfully with material during class
2. Experiential learning – learning through doing, building, testing, iterating
3. Technology-aided learning – simulation, CAD, virtual labs, data acquisition
4. Flipped-classroom-based learning – active use of class time for application
5. Problem-solving and critical-thinking-based learning – authentic engineering challenges

Assessment Practices

Engineering assessment is structured around multiple levels of outcomes that cascade from course-level to programme-level to institutional-level [20], [21]:

Outcome Level	Description	Example
Course Outcomes (COs)	What students can do after a specific course	“Design a bridge truss to meet specified load requirements.”
Program Outcomes (POs)	Graduate attributes across the degree	“Ability to design systems within realistic constraints”
Program Specific Outcomes (PSOs)	Discipline-specific competencies	“Apply civil engineering codes and standards.”
Program Educational Objectives (PEOs)	What graduates achieve 3-5 years post-graduation	“Lead engineering teams in sustainable infrastructure projects”

Assessment methods include:

• Capstone design projects – culminating experiences requiring synthesis of the entire curriculum [21]
• Laboratory reports and practical examinations
• Peer and self-assessment of team contributions
• Portfolios demonstrating progressive mastery
• Standardised disciplinary exams (e.g., Fundamentals of Engineering)

Undergraduate vs Postgraduate Differences

Level	Focus	Pedagogy	Assessment
Undergraduate	Foundational sciences, core engineering principles, design methodology	Flipped classrooms, lab sessions, project-based courses, and industry internships	Problem sets, lab reports, design projects, capstone, standardised exams
Postgraduate	Advanced specialisation, research methods, and original contributions to engineering knowledge	Research seminars, supervised thesis, collaborative industry projects	Thesis/dissertation, journal publications, patent applications, technical presentations

4. Medical Sciences vs Computer Science

Research Process

• Question conceptualisation: Hypothesis-driven, often via PICO (Population, Intervention, Comparison, Outcome). Strong ethical oversight (IRB). Questions about disease aetiology, treatment efficacy, or diagnostics (Rosenberg & Haynes, 2002).
• Aims & objectives: Establish safety, efficacy, sensitivity/specificity, or survival benefit. Often hierarchical: preclinical → Phase I–IV trials [22].
• Methods: Randomised controlled trials (RCTs), cohort studies, case-control studies, and systematic reviews with meta-analysis. Blinding, placebo control, and intention-to-treat analysis are gold standards.
• Example impactful paper: Diabetes Control and Complications Trial (DCCT, 1993) – RCT showing intensive glucose control reduces complications; changed endocrinology.
• Differences with CS: Medical research demands causal inference with high internal validity, can handle small, heterogeneous samples, and is subject to regulatory review. CS often uses observational data (logs, benchmarks) and prioritises predictive accuracy (AUC, F1) over causality, unless explicitly designed to do otherwise.

Medical and Biomedical Sciences Pedagogy: Integration, Simulation, and Competency

Pedagogical Philosophy

Medical and biomedical science education has undergone a significant transformation from discipline-based (anatomy, physiology, biochemistry taught separately) to integrated, systems-based curricula that mirror clinical reasoning [23]. The emphasis is on preparing students for the complexities of real-world practice, not merely knowledge acquisition.

Best-practice pedagogies [23]:

• Case-based learning (CBL): Students analyse patient cases, integrating basic science knowledge with clinical reasoning.
• Team-based learning (TBL): Structured small-group learning with individual and team readiness assurance tests.
• Simulation-based education: High-fidelity manikins, standardised patients, and virtual reality for practising clinical skills without patient risk.
• Research-integrated teaching: Students engage with primary literature and may participate in ongoing research projects.
• Distributed learning: Online and blended approaches for theoretical content, freeing in-person time for skills practice.

Assessment Practices

Medical education is moving toward competency-based assessment that evaluates students’ ability to perform authentic clinical tasks.

Key assessment modalities:

• Objective Structured Clinical Examinations (OSCEs): Students rotate through stations performing specific clinical tasks with standardised patients.
• Written examinations: Often multiple-choice questions designed to test clinical reasoning, not just recall (e.g., USMLE Step exams).
• Portfolio-based assessment: Collecting evidence of clinical encounters, reflective writing, and feedback from supervisors.
• Direct observation of procedural skills (DOPS): Supervisors assess students as they perform procedures.
• Capstone experiences that integrate knowledge, skills, and professional values [23].

Formative Assessment

Frequent, low-stakes feedback is essential, particularly for clinical skills. Many programmes use spaced repetition and low-stakes quizzes to reinforce foundational knowledge.

Undergraduate vs Postgraduate Differences

Level	Focus	Pedagogy	Assessment
Undergraduate (MBBS/BMBS)	Basic medical sciences, clinical skills, and professional formation	Lectures, small-group CBL/TBL, anatomy dissection, clinical simulation, ward placements	Written exams, OSCEs, portfolios, clinical placement evaluations
Postgraduate (residency/fellowship)	Specialised clinical training, research, and leadership	Supervised clinical practice, research projects, morbidity and mortality conferences, and journal clubs	In-training examinations, board certification exams, and thesis (for research pathways)

5. Basic Sciences (Biology, Chemistry, Physics) vs Computer Science

Research Process

• Question conceptualisation: Rooted in natural phenomena – often reductionist (e.g., “What is the structure of this protein?”” or fundamental laws (“What is the Higgs mechanism?”). Hypotheses are falsifiable [24].
• Aims & objectives: Discover laws, mechanisms, or taxonomies; replicate phenomena under controlled conditions.
• Methods: Controlled experiment, systematic observation, instrument measurement (e.g., microscope, spectrometer, particle collider): replication, blinding, and statistical testing (p-values, confidence intervals).
• Example impactful paper: Watson & Crick (1953) – “Molecular Structure of Nucleic Acids” – used X-ray diffraction data and model building; changed biology [25].
• Differences with CS: Basic sciences study pre-existing natural systems; CS studies artefacts (programs, algorithms, systems) created by humans. CS can modify its “universe” arbitrarily; physics cannot change gravity. CS also uses synthetic benchmarks and simulation as first-class methods, not just proxies.

Basic Sciences (Biology, Chemistry, Physics): Laboratory-Integrated Inquiry

Pedagogical Philosophy

The centrality of laboratory work distinguishes the basic sciences as both a pedagogical method and an assessment focus. Best practice integrates theoretical instruction with hands-on investigation, teaching students to think and act like scientists [26].

Best-practice pedagogies:

• Process-oriented guided inquiry learning (POGIL): Particularly strong in chemistry, where students discover concepts through structured team activities.
• Course-based undergraduate research experiences (CUREs): Entire classes conduct original research, producing novel findings.
• Flipped classrooms with pre-laboratory preparation.
• Peer-led team learning (PLTL): Upper-level students facilitate small-group problem-solving sessions.
• Computational and data science integration: Teaching programming and data analysis within disciplinary contexts.

Assessment Practices

Assessment in the basic sciences focuses on both product (correct results) and process (scientific reasoning, lab technique, data analysis) [26].

Key assessment types:

• Laboratory notebooks: Assessed for completeness, accuracy, and quality of reasoning
• Lab reports and scientific papers: Formal write-ups following disciplinary conventions
• Practical examinations: Students perform experiments or operate instruments under observation
• Research proposals: Students design original experiments
• Poster and oral presentations of findings

Formative Assessment

Frequent, accurate feedback is particularly important for laboratory skills, where incorrect techniques can propagate through an experiment. The emphasis is on reinforcing productive laboratory habits and providing specific guidance for improvement [26].

Undergraduate vs Postgraduate Differences

Level	Focus	Pedagogy	Assessment
Undergraduate	Fundamental concepts, lab techniques, and scientific writing	Lectures, laboratory sessions, recitations, and CUREs for advanced students	Lab reports, practical exams, problem sets, midterm/final exams, senior thesis
Postgraduate (PhD)	Original research, advanced methods, scientific communication	Research group meetings, journal clubs, methods courses, and supervised dissertation	Qualifying exams, dissertation, first-author publications, grant writing, and teaching experience

6. Computer Science Research Process

Computer science is methodologically diverse, often borrowing from mathematics, engineering, and experimental science. The core is artefact design and evaluation [1].

Research Conceptualisation

• Types of research questions:
  1. Algorithmic: “What is the most efficient algorithm for problem X?”
  2. Systems: “How to design a distributed system with property Y?”
  3. Empirical: “How do users interact with interface Z?”
  4. Theoretical: “What is the computational complexity of problem W?”

Aims & objectives

• The typical computer science research process involves designing a novel algorithm, proving its properties, implementing it, and evaluating it against baselines. The 2017 Transformer paper [27] applied this template. It introduced a core innovation – attention‑only sequence processing. However, the paper itself did not exhaustively evaluate all possible variations. Subsequent research filled these knowledge gaps by applying the Transformer at different scales, in new domains, and with new interpretations, leading to LLMs of today’s scale. This pattern reveals that a contribution need not be a new method; it can also be a new evaluation of a known method with novel interpretations. 

Method plan – Tripartite structure (Easterbrook et al., 2008) [28]

Paradigm	Method	Evidence
Theoretical	Axiomatic proof, complexity analysis	Formal proof, asymptotic bounds
Empirical	Controlled experiment, case study, survey	Statistical significance, effect size
Engineering	Prototype, benchmarking, simulation	Runtime, memory, throughput, accuracy

• Reproducibility is a growing mandate: code, data, and environment (e.g., Docker, Papers with Code).
• Common pitfalls: Overfitting to benchmarks, lack of external validity, and ignoring negative results.

Example impactful CS paper following this process

Vaswani et al. (2017) – “Attention Is All You Need” [27]

Question: How to perform sequence transduction (e.g., machine translation) without recurrent or convolutional neural networks, which are inherently sequential and difficult to parallelise?

Aims: Design a novel network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Demonstrate superior quality, greater parallelizability, and significantly reduced training time compared to existing models.

Method: Theoretical + engineering + empirical – proposed the Transformer architecture with scaled dot-product attention, multi-head attention, and positional encoding. Implemented the model and evaluated on:

• WMT 2014 English-to-German translation: Achieved 28.4 BLEU, improving over existing best results (including ensembles) by over 2 BLEU
• WMT 2014 English-to-French translation: Established a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs – a small fraction of the training costs of previous best models
• English constituency parsing: Demonstrated strong generalisation to other tasks with both large and limited training data

Impact: Fundamentally changed deep learning and artificial intelligence. The Transformer architecture became the foundation for:

• Large Language Models (LLMs): BERT, GPT (all versions), T5, Llama, and virtually all modern LLMs
• Vision Transformers (ViT): Extended to computer vision
• Multimodal AI: Models like DALL-E, CLIP, and Flamingo
• All sequence modelling tasks: From protein folding (AlphaFold) to music generation

As of 2025, the paper has been cited more than 173,000 times, placing it among the top ten most-cited papers of the 21st century.

Dean & Ghemawat (2008) – “MapReduce: Simplified Data Processing on Large Clusters” [29]

• Question: How to process large-scale data on commodity hardware reliably?
• Aims: Design a programming model and runtime that hides parallelism, fault tolerance, and distribution.
• Method: Engineering + empirical – built system, ran benchmarks (Grep, Sort, WordCount) on thousands of machines, measured speedup and fault recovery.
• Impact: Changed big data, cloud computing, and inspired Hadoop, Spark, etc.

Research Methods Differences from all other disciplines summarised:

• Artefact-centric: CS research produces software, protocols, or formal models – not just knowledge about nature or society.
• Hybrid methodology: Can switch between proof, experiment, and build–evaluate in one paper.
• Benchmark culture: Standardised datasets (ImageNet, SQuAD, SPEC) act as “experimental apparatus” shared globally.
• Fast iteration: No physical prototyping costs; change code, re-run.
• Null hypothesis testing is less common; instead, “improves over baseline” is more common.

Computer Science Pedagogy: Competency-Based, Project-Driven, and Formative

Pedagogical Philosophy

Computer science education has evolved from lecture-based programming instruction to competency-based, project-driven learning that emphasises what students can do, not just what they know [30], [31]. The field addresses a unique challenge: students enter with wildly different prior programming experience, requiring flexible pacing and multiple entry points.

Best-practice pedagogies:

• Competency-based assessment (CBA): Students progress upon demonstrating mastery of specific competencies, not seat time. This approach narrows the gap for less experienced students while still challenging advanced learners [30].
• Constructionism (Papert): Students learn by building shareable artefacts (e.g., programs, websites, apps) that have meaning to them [30].
• Project-first approaches (also called “project-led” or “problem-based”) engage students with authentic programming challenges from the beginning, learning language features as needed [30].
• Active blended learning: Combining digital and physical tools to create engaging, interactive learning experiences [30], [31].
• Peer instruction and pair programming: Structured collaboration that mirrors professional practice.

Assessment Practices

CS assessment is distinctive in its emphasis on iterative, multi-channel feedback that mirrors professional software development [30], [31].

Key assessment types:

Assessment Type	Description	Best Practice
Automated programming exercises	Code is submitted and automatically tested against unit tests	Provides immediate, low-stakes feedback; allows multiple resubmissions [30]
Code reviews	Students review each other’s code with structured rubrics	Develops professional communication and critical reading skills [30], [31]
Portfolio assessment	Collection of projects demonstrating progressive mastery	Emphasises growth and final product quality over single assessments [30]
Live coding exams	Students write code under observation	Assesses authentic problem-solving under pressure
Group projects with individual accountability	Team-based development with mechanisms for assessing individual contribution	Mirrors industry practice while preventing free-riding [30], [31]

Formative Assessment in CS

Course-wide, iterative formative assessment is a growing best practice. Rather than isolated assessments within individual units, a programme-wide approach provides:

• Multiple feedback channels (instructor, peer, automated, end-user)
• Progressive skill scaffolding across semesters
• Early identification of struggling students for intervention [30], [31]

The competency-based approach particularly benefits introductory programming, where students can work at different paces while their progress is monitored in real-time [30].

Undergraduate vs Postgraduate Differences

Level	Focus	Pedagogy	Assessment
Undergraduate	Programming fundamentals, data structures, algorithms, software design, systems	Lecture-labs, project-based courses, hackathons, internships, and team projects	Programming assignments, exams with coding components, group projects, capstone, and code reviews
Postgraduate	Advanced specialisation (AI, systems, theory), research methods, original contribution	Research seminars, supervised thesis, collaborative projects with industry, and reading groups	Thesis/dissertation, conference/journal publications, reproducibility packages, teaching assistantship

Unifying Principles Across Disciplines Pedagogy

Despite their differences, several best-practice principles transcend disciplinary boundaries:

1. Constructive Alignment (Biggs & Tang)

All disciplines are moving toward alignment between learning outcomes, teaching activities, and assessment tasks. What students are taught should be precisely what they are assessed on, and assessments should require the cognitive processes articulated in outcomes [32].

2. Authentic Assessment

Across fields, the gold standard is an assessment that mirrors real-world professional practice:

• Social sciences: policy briefs, not just essays
• Engineering: design projects, not just problem sets
• Medicine: OSCEs with standardised patients
• CS: code reviews and deployed projects

3. Formative Assessment and Feedback

The shift from summative-only to formative-dominant assessment is universal. Frequent, low-stakes, actionable feedback improves learning more than any single high-stakes examination.

4. Active and Student-Centred Learning

Passive lectures are retreating across all disciplines in favour of active learning, PBL, flipped classrooms, and collaborative inquiry.

5. Outcome-Based Education (OBE)

Particularly strong in engineering and medical education, OBE is influencing all fields: teaching and assessment are designed backwards from clearly articulated graduate outcomes.

6. Integration of Research and Teaching

At postgraduate (and increasingly undergraduate) levels, students learn by doing research—not just reading about it. Course-based research experiences (CUREs) are expanding beyond the basic sciences.

Summary 

Discipline	Core Pedagogy	Signature Assessment	Unifying Framework
Social Sciences	Critical pedagogy, PBL, community-engaged	Authentic tasks (policy briefs, portfolios)	Authentic pedagogy
Mathematics	IBL, POGIL, flipped classroom	Proofs, problem sets, and oral exams	Conceptual understanding
Engineering	Active, experiential, technology-aided, PBL	Capstone, labs, outcome-based rubrics	Outcome-Based Education
Medical Sciences	Case-based, team-based, simulation	OSCEs, portfolios, DOPS	Competency-based education
Basic Sciences	Lab-integrated, CUREs, POGIL	Lab notebooks, practical exams, proposals	Inquiry-based learning
Computer Science	Competency-based, constructionism, project-first	Automated testing, code reviews, and portfolios	Competency/Project-based

This analysis demonstrates that each discipline has developed pedagogies and assessments suited to its epistemology and professional context, while also converging on common principles: authenticity, alignment, active learning, and formative feedback. The challenge for modern higher education is to implement these evidence-based practices at scale across diverse disciplinary cultures. For computer science, the practical implication is that interdisciplinary collaboration should be built on strong computational competencies: implementation, testing, debugging, benchmarking, reproducibility, and iterative design. 

References

[1] M. Tedre, The Science of Computing: Shaping a Discipline. CRC Press, 2014. 

[2] R. McElreath, R. Boyd, and P. J. Richerson, ‘Shared Norms and the Evolution of Ethnic Markers’, Curr. Anthropol., vol. 44, no. 1, pp. 122–130, Feb. 2003, doi: 10.1086/345689. 

[3] U. Nations, Universal Declaration of Human Rights. 1948. Accessed: May 29, 2026. [Online]. Available: https://www.un.org/en/about-us/universal-declaration-of-human-rights 

[4] Congress, Civil Rights Act of 1964. 2015. Accessed: May 29, 2026. [Online]. Available: https://www.govinfo.gov/app/details/https%3A%2F%2Fwww.govinfo.gov%2Fapp%2Fdetails%2FCOMPS-342 

[5] U. Government, Equality Act 2010. Accessed: May 29, 2026. [Online]. Available: https://www.legislation.gov.uk/ukpga/2010/15/contents 

[6] Charter of Fundamental Rights of the European Union. 2012. Accessed: May 29, 2026. [Online]. Available: https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:C2012/326/02 

[7] M. Helal, ‘Shared Norms and Freedom, Equality, Ethical and Legal Behaviours’. Accessed: May 29, 2026. [Online]. Available: https://manalhelal.com/2023/06/shared-norms-and-freedom-equality-ethical-and-legal-behaviours/ 

[8] P. E. Griffiths, ‘Why the “interdisciplinary” push in universities is actually a dangerous antidisciplinary trend’, The Conversation. Accessed: May 29, 2026. [Online]. Available: https://theconversation.com/why-the-interdisciplinary-push-in-universities-is-actually-a-dangerous-antidisciplinary-trend-175511 

[9] G. Squires, ‘Interdisciplinarity in Higher Education in the United Kingdom’, Eur. J. Educ., vol. 27, no. 3, p. 201, 1992, doi: 10.2307/1503449. 

[10] Y. Hong, B. Kim, J. Jeon, and L. Kim, ‘Has higher education become more interdisciplinary? a longitudinal analysis of syllabi using natural language processing’, Humanit. Soc. Sci. Commun., vol. 12, no. 1, p. 1841, Nov. 2025, doi: 10.1057/s41599-025-06126-7. 

[11] M. Crotty, The foundations of social research: meaning and perspective in the research process. London ; Thousand Oaks, Calif: Sage Publications, 1998. 

[12] C. Geertz, The interpretation of cultures: selected essays. New York: Basic Books, 1973. 

[13] M. S. Granovetter, ‘The Strength of Weak Ties’, Am. J. Sociol., vol. 78, no. 6, pp. 1360–1380, 1973. 

[14] C. O. Alordiah and R. Arisi, ‘Toward a Critical Pedagogy of Assessment Integrating Social Studies Education: The Proposed Socio-Critical Assessment Integration Framework (SCAIF)’, J. Res. Initiat., vol. 8, no. 5, 2025, Accessed: May 29, 2026. [Online]. Available: https://eric.ed.gov/?id=EJ1472189 

[15] J. Gore, J. Ladwig, W. Elsworth, H. Ellis, R. Parkes, and T. Griffiths, ‘Quality assessment: linking assessment tasks and teaching outcomes in the social sciences’, Open Research Newcastle, report, Jan. 2009. Accessed: May 29, 2026. [Online]. Available: https://openresearch.newcastle.edu.au/articles/report/Quality_assessment_linking_assessment_tasks_and_teaching_outcomes_in_the_social_sciences/28960070/1 

[16] K. Gödel, On formally undecidable propositions of Principia mathematica and related systems. New York: Dover Publications, 1931. 

[17] ‘Notes Volume 89 – MAA Instructional Practices Guide – Mathematical Association of America’. Accessed: May 29, 2026. [Online]. Available: https://maa.org/resource/notes-volume-89-maa-instructional-practices-guide/ 

[18] J. D. Bostic, E. E. Krupa, J. C. Shih, and Taylor & Francis, Eds, Assessment in mathematics education contexts: theoretical frameworks and new directions. in Routledge research in education, no. 43. New York, NY: Routledge, 2019. 

[19] S. D. Sheppard, K. Macatangay, A. Colby, and W. M. Sullivan, ‘Educating Engineers: Designing for the Future of the Field’, Carnegie Foundation for the Advancement of Teaching, 2008. Accessed: May 29, 2026. [Online]. Available: https://eric.ed.gov/?id=ED504076 

[20] A. Mishra and S. S. Singh, ‘Smart pedagogical learning and assessment methodologies—outcome based education: A novel quality assurance framework with attainment computation for Indian engineering higher education programs’, J. Comput. Educ., vol. 12, no. 4, pp. 1231–1282, Dec. 2025, doi: 10.1007/s40692-024-00344-9. 

[21] R. M. Felder, J. Spurlin, S. A. Rajala, and J. P. Lavelle, Designing Better Engineering Education Through Assessment: A Practical Resource for Faculty and Department Chairs on Using Assessment and ABET Criteria to Improve Student Learning, 1st edn. New York: Routledge, 2023. doi: 10.4324/9781003444107. 

[22] D. L. Sackett, W. M. C. Rosenberg, J. A. M. Gray, R. B. Haynes, and W. S. Richardson, ‘Evidence based medicine: what it is and what it isn’t’, BMJ, vol. 312, no. 7023, pp. 71–72, Jan. 1996, doi: 10.1136/bmj.312.7023.71. 

[23] D. Johnson, Ed., Pedagogies of Biomedical Science: A Holistic Approach to Integrating Pedagogy Across the Curriculum. London: Routledge, 2024. doi: 10.4324/9781003383994. 

[24] K. R. Popper, Conjectures and refutations: the growth of scientific knowledge. in Routledge classics. London ; New York: Routledge, 2002. 

[25] J. D. Watson and F. H. C. Crick, ‘Molecular Structure of Nucleic Acids: A Structure for Deoxyribose Nucleic Acid’, Nature, vol. 171, no. 4356, pp. 737–738, Apr. 1953, doi: 10.1038/171737a0. 

[26] R. L. Doran, Science educator’s guide to laboratory assessment, 2nd enlarged ed. Arlington, Va.: National Science Teachers Association, 2002. 

[27] A. Vaswani et al., ‘Attention Is All You Need’, in NeurIPS Proceedings, LONG BEACH CA, Dec. 2017. doi: https://dl.acm.org/doi/10.5555/3295222.3295349. 

[28] S. Easterbrook, J. Singer, M.-A. Storey, and D. Damian, ‘Selecting Empirical Methods for Software Engineering Research’, in Guide to Advanced Empirical Software Engineering, F. Shull, J. Singer, and D. I. K. Sjøberg, Eds, London: Springer, 2008, pp. 285–311. doi: 10.1007/978-1-84800-044-5_11. 

[29] J. Dean and S. Ghemawat, ‘MapReduce: simplified data processing on large clusters’, Commun. ACM, vol. 51, no. 1, pp. 107–113, Jan. 2008, doi: 10.1145/1327452.1327492. 

[30] K. Androutsopoulos, M. Heeney, S. Smith, and M. A. Ticar, ‘Lessons from Adopting a Competency-based Assessment Approach for an Introductory Programming Module’, in Proceedings of the 30th ACM Conference on Innovation and Technology in Computer Science Education V. 2, Nijmegen Netherlands: ACM, Jun. 2025, pp. 743–744. doi: 10.1145/3724389.3731257. 

[31] I. Benardis, A. Hayes, and J. H. Davenport, ‘Beyond the unit: A course-wide, iterative formative assessment and feedback framework for enhancing learning and employability skills in computer science education’, in Formative Assessment and Feedback in Post-Digital Learning Environments, London: Taylor and Francis/ Balkema, 2025, pp. 186–194. doi: 10.4324/9781003360254-24. 

[32] J. B. Biggs and C. S. Tang, Teaching for quality learning at university: what the student does, 4th edition. in SRHE and Open University Press Imprint. Maidenhead: McGraw-Hill/Society for Research into Higher Education/Open University Press, 2011.