Manifold Grace

RESEARCH & KNOWLEDGE · ADDITIVE PREPARATION RESOURCE · 11 NOVEMBER 2026

MA → Dissertation → Viva → Doctoral Research: Critical Polishing & Self-Assessment

A source-preserving preparation space for the MA AI, Ethics and Society (2024–25), alongside a separate 2026 landscape review and prospective doctoral work. This is not a replacement for the approved syllabus, proof of qualification completion, or an official university examination. Existing research is substantive; new candidate studies are marked separately.

Programme baseline: 180 credits at Level 7: two 30-credit core modules, 60 elective credits, and a 60-credit independent dissertation. For the two-year October part-time route, the 2024–25 specification places the two taught core modules in Year 1 and the dissertation plus remaining options in Year 2. The dissertation module describes a 12,000-word submission. Original approved specification.

Jump to preparation area

Learning outcomesCore & option modulesReading inventoryInteractive mastery auditDissertation/viva50 doctoral topicsLearning-outcome projectsReferences

Nine original programme learning outcomes

Faithful substantive summaries of the nine separately numbered outcomes, with the official document retained as authority. For quoted exact language, consult page 6 of the official specification.

  1. LO1 — Explain AI principles, technological foundations and societal impacts
  2. LO2 — Evaluate ethical and societal implications across technology, law, policy, philosophy and business
  3. LO3 — Critically analyse AI case studies and propose viable responses
  4. LO4 — Apply ethical theories/frameworks to practical AI dilemmas
  5. LO5 — Develop and critique policy and regulation with philosophical and technical reasoning
  6. LO6 — Discuss and debate complex ethical questions clearly and coherently
  7. LO7 — Communicate AI ethics to policymakers, technologists and the public
  8. LO8 — Conduct independent research culminating in a well-argued contemporary thesis
  9. LO9 — Collaborate across academic and professional sectors on AI governance strategies

Course structure without retrospective substitution

Year 1 and Year 2

StageConfirmed requirementCredit basisPreparation focus
Part-time Year 1Ethics of Technology and AI; History and Foundations of AI; selected optionsTwo core modules, 30 + 30; Year 1 options limited by original rulesEthics and AI histories, foundational debates, cases and optional study
Part-time Year 2Dissertation in AI, Ethics and Society; remaining optionsDissertation 60; remaining option credits to 180 totalIndependent thesis, debate mastery, research method and evidence defence

Optional-module audit: no original option omitted

The first ten rows reproduce the titles, codes and credits of all ten distinct options found in the 2024–25 specification. The last three are additions on the 2026 course listing, explicitly not assumed compulsory for 2024–25.

CodeOption moduleCreditsVersion/status
SC03020S7Philosophy of Artificial Intelligence: Minds, Machines and Meaning302024–25 option
ARMC247S7Managing Digital Media302024–25 option
LALW059S7Algorithms and the Law302024–25 option
BUCI081H7Demystifying Computing with Python152024–25 option
BUCI080H7Analytic Tools for Data Science152024–25 option
BUMN154H7Managing Emerging Digital Technologies152024–25 option
BUMN176H7Introduction to Analytics and Business152024–25 option
BUMN178H7Big Data Management152024–25 option
BUMN179H7Data Visualization and Communication152024–25 option
SC04023S7Critical and Creative AI302024–25 option (full-time list; not present in extracted October part-time Year 1 option subset)
2026AI, Language and Social Justice—Advertised 2026; not retrospectively required
2026GeoAI for Social and Urban Applications—Advertised 2026; not retrospectively required
2026Regulating Technology—Advertised 2026; not retrospectively required

History and Foundations: ten-week knowledge map

Week 1: Thinking machines

AI as essentially contested; weak/narrow AI, AGI and ASI; McCarthy, Rosenblatt, Minsky; Turing versus Searle

Evidence activity: identify the primary argument, timeline anchor, opposing interpretation and one contemporary case.

Week 2: Cybernetics and grounded cognition

Signal processing, feedback and control; Barsalou/Wiener; bodily grounding and socially constructed intelligence

Evidence activity: identify the primary argument, timeline anchor, opposing interpretation and one contemporary case.

Week 3: Expert systems to ChatGPT

Symbolic and probabilistic systems; encoded expertise; AI winters; anthropomorphic and parasocial design

Evidence activity: identify the primary argument, timeline anchor, opposing interpretation and one contemporary case.

Week 4: Probability and prediction

Historical probability; supervised learning/deep-learning differences; forecasting and Rescher's typology

Evidence activity: identify the primary argument, timeline anchor, opposing interpretation and one contemporary case.

Week 5: Data as infrastructure

Datasets, annotation, classification, data work, labour, consent, extractive political economy

Evidence activity: identify the primary argument, timeline anchor, opposing interpretation and one contemporary case.

Week 6: Feedback and model collapse

Synthetic data, feedback loops, recursive degradation and fairness amplification

Evidence activity: identify the primary argument, timeline anchor, opposing interpretation and one contemporary case.

Week 7: AI, power and the state

Welfare and policing, public-service administration, Eubanks, Benjamin, Winner

Evidence activity: identify the primary argument, timeline anchor, opposing interpretation and one contemporary case.

Week 8: Fairness and counter-imaginaries

Bias and fairness measurement, technical interventions, systemic reforms, alternative institutional imaginaries

Evidence activity: identify the primary argument, timeline anchor, opposing interpretation and one contemporary case.

Week 9: Design and accountability

Affordances, explanation, interpretability, transparency, responsibility and design justice

Evidence activity: identify the primary argument, timeline anchor, opposing interpretation and one contemporary case.

Week 10: Materiality of AI

Data-centre energy/water, compute, extractive supply chains, labour and planetary costs

Evidence activity: identify the primary argument, timeline anchor, opposing interpretation and one contemporary case.

Reading and source audit

Selected named readings identified in the supplied teaching documents, with the publication year and assigned reading status where confirmed. This is an expandable register, not a claim that these 22 records exhaust the complete weekly bibliography or all other modules. Where the version/reprint differs from the first publication, check the specific syllabus passage.

Author(s)WorkYearWeekAssignment status
TuringComputing Machinery and Intelligence19501Primary
SearleMinds, Brains, and Programs19801Primary
MinskySteps Toward Artificial Intelligence19611Additional
McCarthyPrograms with Common Sense19591Additional
HalpernBeautiful Data20152Primary, opening chapters
BarsalouGrounded Cognition20082Reading-list companion
WienerCybernetics19482Additional (1961 edition passage)
FoggPersuasive Technology20033Reading-list text
Jiang, Gradus & RosselliniSupervised machine learning: A brief primer20204Primary
RescherPredicting the Future19984Primary, chapters 7–8
CrawfordAtlas of AI20215Primary, data chapter; earth chapter in Week 10
Muldoon et al.A typology of artificial intelligence data work20245Primary
Shumailov et al.The Curse of Recursion20246Primary
Wyllie, Shumailov & PapernotFairness Feedback Loops20246Primary
EubanksAutomating Inequality20187Primary, chapter 3
BenjaminRace After Technology20197Primary, chapter 3
WinnerDo Artifacts Have Politics?19807Additional
Kazansky & MilanBodies not templates20218Primary
FerraraFairness and Bias in Artificial Intelligence20238Primary
MitchellArtificial Intelligence: A Guide for Thinking Humans2019BackgroundSuggested background
BodenAI: Its Nature and Its Future2016BackgroundSuggested background
Cave, Dihal & DillonAI Narratives2020BackgroundSuggested background

Reading protocol: extract thesis, key distinctions, central evidence, strongest criticism, historical context, original publication date, later revised editions and relevance to your research. Preserve uncertain or unverified citations as such.

Ten knowledge domains — deepening register

These domains encompass books, topics, history, computing, philosophy, terminology, political constructs, statistics, modern events and research. They supplement rather than diminish the original syllabus.

1. Books and literature

Primary and supplemental authors; original dates; text editions; positions; criticism; reading not yet confirmed

Preparation standard: define the concept, locate primary evidence, record date/version, formulate the best objection and show application; attach source or demonstration before claiming mastery.

2. History and chronology

Computing and AI history; first publication years; revision history; conditions for causal historical claims

Preparation standard: define the concept, locate primary evidence, record date/version, formulate the best objection and show application; attach source or demonstration before claiming mastery.

3. Computing and AI

Algorithms, symbolic methods, ML, deep learning, transformers, RAG, agents, infrastructure, security, evaluation

Preparation standard: define the concept, locate primary evidence, record date/version, formulate the best objection and show application; attach source or demonstration before claiming mastery.

4. Philosophical theories

Metaphysics, mind, epistemology, ethics, philosophy of science, political philosophy and technology philosophy

Preparation standard: define the concept, locate primary evidence, record date/version, formulate the best objection and show application; attach source or demonstration before claiming mastery.

5. Terminology and concepts

Definitions, rival uses, concept boundaries, necessary/sufficient conditions and counterexamples

Preparation standard: define the concept, locate primary evidence, record date/version, formulate the best objection and show application; attach source or demonstration before claiming mastery.

6. Political constructs

State structure, administrative accountability, rights, procedural justice, ideology, institutions and global AI governance

Preparation standard: define the concept, locate primary evidence, record date/version, formulate the best objection and show application; attach source or demonstration before claiming mastery.

7. Statistics and methods

Sampling, inference, precision/recall, calibration, regressions, causality, fairness tests, mixed methods

Preparation standard: define the concept, locate primary evidence, record date/version, formulate the best objection and show application; attach source or demonstration before claiming mastery.

8. 2026 research landscape

As-of dated policy, institution and technology developments, competing positions and source confidence

Preparation standard: define the concept, locate primary evidence, record date/version, formulate the best objection and show application; attach source or demonstration before claiming mastery.

9. Dissertation and viva

Thesis question, scope, literature gap, methodology, conceptual precision, objections, examiner defence

Preparation standard: define the concept, locate primary evidence, record date/version, formulate the best objection and show application; attach source or demonstration before claiming mastery.

10. PhD research design

Original contribution, strongest prior art, falsification, replicability, supervision fit and publication pathway

Preparation standard: define the concept, locate primary evidence, record date/version, formulate the best objection and show application; attach source or demonstration before claiming mastery.

Interactive reasoning and knowledge audit — 50 questions

Five questions in each of ten domains, with dated facts, conceptual distinctions, methodological inference and programme-specific requirements. Answers are not transmitted to a server. Each category reports a percentage of available points, not a genuine population percentile rank; a population percentile would require a validated comparison cohort. Questions are for revision and cannot certify doctoral or degree-level competence.

Dissertation, thesis and viva-style defence

Existing research draft: Ethical Personhood, Bias, and Systemic Moral Accountability in AI, including human bias, unethical delegation, responsibility and value-embedding in socio-technical infrastructures. The uploaded dissertation draft establishes that this is already a developed research direction; the page does not imply it remains unstarted.

Master's requirement: a 60-credit dissertation, conventionally 12,000 words in the referenced module description. Doctoral standard: a defensible original contribution assessed through an institution-specific doctoral process. A viva-style rehearsal for MA work is useful, but does not imply that a formal viva was required by the original module.

WorkstreamEvidence to prepareViva-style questions
Thesis and scopeOne-sentence claim; precise contribution; bounded population, time, jurisdiction and AI systemWhat precisely are you arguing? What is not being claimed?
Literature & noveltyStrongest prior work, date-ordered evidence, overlapping claims, falsifiable differenceWhat has already been established? What survives the overlap audit?
Philosophical reasoningExplicit premises, counterarguments, moral and political distinctionsWhy this theory? Which criticism is strongest?
MethodologyResearch design, sampling, operational definitions, validity and reproducibilityWhy this method rather than its closest alternative?
Results & uncertaintySources, versioned findings, statistical or textual evidence and limitationsWhat observation would disconfirm your conclusion?
Research integrityCitation, intellectual provenance, AI assistance disclosures, ethics and data rightsCan the result be independently checked and responsibly reproduced?
Contributions & implicationsWhat can be adopted, revised, studied or challengedWhose decisions would legitimately change as a result?
Doctoral progressionResearch question and candidate supervisors' expertise, funding/entry routes and institutional requirementsWhat new generalisable knowledge could a PhD establish?
Dissertation type possibilities

Normative-philosophical thesis; conceptual analysis; history of ideas; legal/doctrinal analysis; policy evaluation; critical socio-technical case study; qualitative interview/ethnographic work (with ethics approval); quantitative observational study; experimental human–AI evaluation; mixed-method investigation; design science and evaluated prototype; reproducibility/benchmark study; comparative institutional analysis; systematic review with a clearly defined contribution; article-based thesis where the doctoral institution permits it. These are possible designs, not all requirements of the MA.

Top 50 doctoral topic candidates — grouped, not ranked

Fifty researchable candidate directions grounded in the established interests and document themes available here. Status identifies overlapping existing directions or a new candidate; it does not certify originality or imply that existing papers are unfinished. All 50 require prior-art and supervision-fit checks before a doctoral proposal is finalised. A decade-long independent audit of LinkedIn posts, every past conversation and all supervisors' publications has not been completed.

No.Research areaPrecise topic candidateRelationship to current work
1Evidence verificationCross-institutional claim-to-evidence coverage benchmarksExisting direction: evidence-verification-across-sectors
2Evidence verificationContradiction-aware synthesis across legal and public datasetsExisting programme: contradiction detection
3Evidence verificationHistorical as-of-date verification for changing institutional claimsExisting proposal: historical retrieval
4Evidence verificationContributor provenance and attribution in human–AI researchExisting contributor-provenance initiative
5Evidence verificationOpen audit trails for corrected AI-generated adviceRelated to evidence integrity
6Institutional accessWhen do extra qualifications improve validity versus restrict access?New flagged investigation
7Institutional accessPortfolio-based recognition of prior learning in AI careersRelated to existing access research
8Institutional accessAI-supported auditing of educational appeals and process consistencyRelated to existing access research
9Institutional accessDisability adjustments and performance validity in AI-era assessmentsRelated to existing access research
10Institutional accessCorporate certification diffusion from employers to universitiesNew flagged investigation
11Agent assuranceContinuous assurance of agentic AI software architecturesExisting agent assurance concept
12Agent assuranceDetecting unsupported autonomy claims in enterprise AI agentsRelated to existing AI evaluation
13Agent assuranceCross-model disagreement as a risk signal in automated workflowsRelated to existing comparison work
14Agent assuranceHuman override usability and accountability in agent workflowsRelated to existing agent oversight
15Agent assuranceReproducible red-teaming for tool-using public-sector agentsCandidate extension
16AI fairnessFairness under recursive synthetic-data trainingSyllabus-aligned extension
17AI fairnessInstitutional versus statistical definitions of fairness in welfare automationSyllabus-aligned extension
18AI fairnessValue-sensitive evaluation of public-benefit algorithmic classificationsRelated to ethics and bias work
19AI fairnessBias amplification across linked administrative decisionsCandidate extension
20AI fairnessProcedural contestability as an AI fairness outcomeRelated to oversight research
21Philosophy of AIOperational tests for warranted machine understanding claimsHistory module extension
22Philosophy of AIThe limits of functionalism in AI responsibility assignmentRelated to dissertation philosophy
23Philosophy of AIEpistemic humility as a measurable property of AI agentsExisting programme direction
24Philosophy of AIBeyond benchmark accuracy: AI uncertainty and self-knowledgeExisting programme direction
25Philosophy of AIHuman values as constraints in socio-technical AI infrastructuresExisting dissertation theme
26Public systemsCoverage gaps between official statistics and media attentionExisting societal intelligence extension
27Public systemsAgency-level accountability mapping for UK public AI deploymentRelated to public-systems directory
28Public systemsInstitutional contradiction detection in public-service guidanceExisting evidence-verification direction
29Public systemsAccessible explanation of algorithmic public decisionsRelated to explainability work
30Public systemsMeasuring human review in automated public-service refusalsCandidate extension
31Economic and labourAI workforce reskilling standards versus competence evidenceRelated to skills and access
32Economic and labourSkills taxonomy drift in AI-related recruitmentRelated to competence mapping
33Economic and labourHuman capital and credential signalling under AI automationCandidate extension
34Economic and labourResponsible AI adoption in small organisations without dedicated compliance teamsCandidate extension
35Economic and labourComparative costs of human–AI review architecturesCandidate extension
36Media and cultureSynthetic content feedback loops and public knowledge diversityHistory syllabus extension
37Media and cultureProvenance verification for AI-generated public informationExisting verification direction
38Media and cultureCross-cultural evaluation of AI explanation accessibilityCandidate extension
39Media and cultureAlgorithmic amplification of neglected community knowledgeRelated to societal intelligence
40Media and cultureHistorical representations of intelligence and contemporary AI narrativesHistory syllabus extension
41Methods and evaluationFalsification-oriented novelty audits for interdisciplinary AI researchExisting gap-first research protocol
42Methods and evaluationBenchmarking claim-to-source traceability at paragraph levelExisting verification programme
43Methods and evaluationCalibrated uncertainty scoring for human–AI joint decisionsCandidate extension
44Methods and evaluationReproducibility standards for live agent demonstrationsRelated to AI portfolio
45Methods and evaluationDetecting construct invalidity in AI fairness metricsCandidate extension
46Infrastructure and environmentData-centre environmental disclosure and verifiabilitySyllabus materiality extension
47Infrastructure and environmentHidden labour in AI data supply chainsSyllabus data-work extension
48Infrastructure and environmentRisk allocation in third-party foundation-model supply chainsCandidate extension
49Infrastructure and environmentPrivacy-preserving evidence access across institutional boundariesRelated to systems architecture
50Infrastructure and environmentGovernance of resource-intensive AI evaluationCandidate extension

Proposal filter for each candidate

Research question → strongest prior art → demonstrable limitation → falsifiable contribution → appropriate method → evidence access → ethics and feasibility → supervisor expertise and institutional fit → thesis type → dissemination strategy. A large topic list is not evidence of fifty separate doctoral-level discoveries.

Summary: curriculum-aligned project portfolio (separate from doctoral topics)

This project register translates core modules, every 2024–25 option, the Year 2 dissertation, and separately labelled 2026 extensions into demonstrable output. LO numbers refer to the original nine outcomes above. Existing work is identified as such; study exercises and candidates are not represented as completed publications.

Mapped curriculumProject or evidence demonstrationLearning outcomesPossible outputStatus
Core ethicsComparative ethical analysis of high-stakes AI cases2,3,4,6Case-study matrix, competing ethical analyses and counterargumentsLearning-outcome study, not new published project
Core historyAI historical chronology and reading provenance atlas1,6,7Timeline with dates, source editions and contested interpretationsStudy extension
AI conceptsInteractive conceptual distinctions examiner1,6,7Terminology quiz and misconception registerStudy extension
Philosophy of AI optionTuring–Searle–embodied cognition argument map1,4,6Premise graph and rebuttal dossierStudy extension
Algorithms and Law optionRegulatory contestability and administrative appeal casebook2,3,5Jurisdiction-specific casebook with date validityStudy extension
Python optionReproducible classification-and-error lab1,3Documented code and confusion matricesDemonstration proposal
Data science optionFairness metric incompatibilities notebook1,2,3Worked fairness calculations and limitationsDemonstration proposal
Big Data optionSynthetic-data feedback and provenance pipeline1,3,5Data lineage and bias-drift testRelated to existing verification research
Data Visualisation optionPublic evidence uncertainty dashboard3,7Accessible charts and source-linked caveatsRelated to existing public-data work
Analytics and Business optionCredential bottleneck evidence dashboard2,3,5Recruitment rules, validity and access indicatorsNew flagged investigation
Emerging technologies optionAgentic deployment governance lifecycle1,5,9Control gates, test plan, review ownersRelated to existing assurance work
Managing Digital Media optionSynthetic media provenance demonstration2,3,7Labelled provenance and correction flowRelated to existing media interests
Critical and Creative AI optionAI authorship and originality test protocol1,4,6Design and evaluation rubric for creative outputStudy extension
2026 advertised new optionAI language and social justice evaluation2,4,7Language fairness and interpretability auditCurrent elective extension, not historical requirement
2026 advertised new optionUrban GeoAI decision transparency map2,3,5Geospatial case-study auditCurrent elective extension
2026 advertised new optionRegulating technology comparative framework2,5,9As-of legal and regulatory matrixCurrent elective extension
DissertationValue embedding and systemic moral accountability thesis2,4,6,8Refined problem, thesis, literature, methodology and objectionsExisting dissertation draft subject
Doctoral methodsGap-first prior-art and falsification protocol6,8Strongest counter-literature, testable gap, stopping rulesExisting research method
Public systemsCross-agency evidence verification prototype3,5,8,9Claim-evidence coverage and contradiction testsExisting programme direction
Professional practiceHuman–AI competency validity pilot3,7,8,9Work samples, scored assessments and fairness analysisCandidate pilot

Primary evidence and dates

  1. Original 2024–25 programme specification (official PDF)
  2. Dissertation module description and outcomes
  3. Current course modules (2026; not retrospective)
  4. QAA Master's Characteristics Statement
  5. QAA qualifications characteristics updates (2026)
  6. OfS/Advance HE AI assessment report (2026)
  7. Related credential-access evidence red flag

The privately supplied History and Foundations of AI lecture syllabus and reading list are summarised here as an additive study guide; no full uploaded teaching file has been republished. The original copies remain unchanged.

Evidence, professional breadth and contribution validation: additive 11 November 2026 review

Preservation rule: This is an append-only companion to the nine historic learning outcomes, original option modules, syllabus, reading register, 50-question self-audit, 50 doctoral candidates and pre-existing project portfolio. It does not alter the source documents or retrospectively apply 2026 standards to the 2024–25 curriculum. All research classifications are provisional unless linked evidence supports completion.

Year 2 dissertation is independent research, not a substitute taught first year

The documented part-time route has 180 credits: two Year 1 core modules of 30 credits each, 60 optional credits spread across the programme, and a 60-credit Year 2 dissertation. The dissertation is an assessed, supervised, independent written research output; it is not automatically a published peer-reviewed paper, nor does an externally published paper automatically satisfy degree assessment. Publication, supervisor approval, submission, research ethics clearance and viva arrangements are distinct processes. A formal oral viva is not established as a mandatory component solely by the materials reviewed; the viva exercises on this page are developmental doctoral-defence preparation.

Relevant knowledge evidenced and additional evidence to review

Knowledge and experienceExisting evidence or statusHow it bears on learning outcomes and doctoral preparationValidation standard
AI ethical inquiry, philosophy and research writing2024–25 submissions, syllabus engagement and drafted dissertation are identifiable in supplied documents.Argumentation, moral responsibility, bias, values and socio-technical consequences (LO2, LO4, LO8, LO9).Preserve drafts, submission receipts, marks and independent methodological review separately; completed study is not identical with an awarded credit.
Social-media thought leadership since 2017Author-reported public discourse and professional commentary; a complete decade-long, post-level audit has not been independently performed.Potential evidence of sustained question formation, competing-perspective analysis, framing and policy interest.Retrieve dated public posts, original URLs, revision histories and responses; evaluate selected claims against current literature and avoid attributing unverified originality.
ChatGPT-supported exploration since 2024Documented conversation-derived research directions and website contributions; not an exhaustive authenticated history of every exchange.Prompted debate, concept synthesis, hypothesis refinement, research taxonomy and falsification design.Compare first dated author statements with AI assistance, later revisions, bibliography and published code; do not equate assisted drafting with independent validation.
Legal advocacy and procedural analysisExperience and correspondence described by the author; private case details, identities and court records deliberately excluded from this public study page.Procedural fairness, administrative reasoning, standards of proof, evidence admissibility, contestability and accountability (LO2, LO6, LO7).Use an anonymised evidence ledger: issue, authority, submission, acknowledgement, outcome, uncertainty. No suggestion that advocacy constitutes accredited legal practice.
Policy revision and communications to representativesSome dated policy proposals and correspondence are separately recorded on the public productivity log.Public policy formation, institutional responsibility, governance of knowledge and reasoned persuasion (LO6, LO7, LO8).Distinguish a proposal submitted, an acknowledgement, substantive feedback, implementation and impact. Do not disclose personal or official recipient names in the public log.
Organisation acknowledgements and external feedbackAuthor reports formal acknowledgements from some organisations and potential further platform review; individual outcomes must be checked against source correspondence.Third-party reception and independent discussion as distinct from academic assessment.Separate automatic receipt, reviewer engagement, endorsement, acceptance and implementation; do not describe future feedback as received.
Business administration, strategy, governance and managementDocumented professional experience and existing business/venture planning, with varying availability of independent corroboration.Process design, incentives, organisational authority, financial and operational trade-offs; useful for AI implementation and governance (LO3, LO6, LO9).Use dated role records, project artefacts, deliverables and measurable outcomes without including private employer disputes.
AV, unified communications, IT systems and quality assuranceProfessional experience described in technology operations, commissioning, latency/jitter, testing, documentation and service support.Concrete analogies for AI system reliability, human-machine interaction, verification protocols, incident response and systems engineering (LO1, LO2, LO9).Where shareable, show redacted test plans, requirements-to-test traceability, troubleshooting analyses and user acceptance evidence; no claim of credentials without records.
Community, health, education and safeguarding practiceInterdisciplinary professional and project themes are identified; formal independent outcome audit is incomplete.Case studies in care, risk, equity, accessibility and responsible decision-making (LO2, LO4, LO7).Protect confidentiality, avoid identifying service users, and distinguish direct practice from proposed applications.
Compliance, transparency, accountability and auditability venture conceptsExisting venture and systems pages document plans and architecture ideas, not necessarily launched businesses.Risk assessment, policy translation, secure deployment, performance assurance, data provenance, business process governance (LO3, LO6, LO9).Map each service to standards, baseline evidence, tests, responsible roles, security controls and an honest launch status.

How to self-validate reasoning rather than self-certify expertise

The existing 50-item quiz measures within-site percentage correct, not population percentile rank or external academic accreditation. Add a separate evidence-based reasoning portfolio for dates, occurrences, sources, frameworks, theories, concepts, policy, quantitative interpretation, uncertainty, case analysis, oral defence and counter-argument handling. A reliable self-assessment distinguishes recollection, unsupported confidence, correct answer, strength of evidence and ability to defend against alternative explanations. For each item record the source and as-of date; where a contradiction arises, record the correction and reason for revision.

Practical self-validation rubric and viva practice

Use four levels: 0 = cannot state accurately; 1 = recognises the term but cannot defend it; 2 = explains accurately with one source and application; 3 = compares competing sources, limitations and a strong objection; 4 = tests a novel prediction or method with reproducible evidence. A percent score equals points earned divided by possible points, not a statistical percentile. For doctoral viva practice: define the claim, locate its novelty relative to the strongest previous research, justify chosen data and methods, consider rival explanations, identify ethical constraints and describe what would falsify the conclusion.

Linked project applications mapped to the course: extended portfolio

The following table adds 35 cross-domain connections to the pre-existing project register without treating them as 35 new unfinished initiatives. Learning outcome numbers refer to the historic nine-outcome summary above. The research source and venture links identify related site materials; a linked page is not itself proof of a finished empirical study.

Study / applied projectEvidence or source linkProgramme fitDocumented statusTest or next evidential question
1. Claim-to-evidence coverage and contradiction APIExisting related resourceLO2, LO4, LO6, LO8, LO9 · explainability, epistemology, evidenceExisting documented proposalTest error-recall, contradiction sensitivity, audit trace and human verification.
2. AI auditing historical contributionsExisting related resourceLO1, LO5, LO8, LO9 · history and attributionPublished historical reviewCorroborate the authorship chronology using dated primary records.
3. Continuous agent architecture assuranceExisting related resourceLO1, LO3, LO6, LO9 · system governancePublished proposalBenchmark against established assurance methods and attack/failure scenarios.
4. Comparative AI assuranceExisting related resourceLO2, LO3, LO7, LO9 · comparison and measurementPublished exploratory analysisRepeat evaluations with reproducible conditions and uncertainty intervals.
5. Evidence integrity across sectorsExisting related resourceLO2, LO4, LO6, LO9 · applied ethicsExisting research directionCompare health, education, law and administration against one measurement protocol.
6. Criminal-justice evidence verificationExisting related resourceLO2, LO4, LO6, LO7 · rights and contestabilityExisting specialist studyEvaluate reversibility, decision appeals, provenance and error impacts.
7. Bias and reliability auditExisting related resourceLO2, LO3, LO4, LO9 · fairnessExisting practical guideQuantify disparate error rates, data coverage and mitigation trade-offs.
8. Fairer socio-technical systemsExisting related resourceLO2, LO4, LO6, LO7 · political philosophyExisting research pageTest distributional effects and who has standing to contest outcomes.
9. Societal intelligence and resilience adviserExisting related resourceLO1, LO2, LO6, LO9 · public-system analysisExisting research programmeTest coverage gaps, trust calibration and escalation to qualified humans.
10. Historical AI and societal intelligenceExisting related resourceLO1, LO5, LO8 · AI historyExisting chronologyValidate independent origins and distinguish prior work from contributions.
11. AI systems atlasExisting related resourceLO1, LO3, LO8 · taxonomyExisting systems comparisonMeasure completeness and reproducibility across domains.
12. National and global frameworks completenessExisting related resourceLO3, LO6, LO9 · regulation and auditingExisting atlasCompare framework control coverage and measurable omissions.
13. Semantic integrity and public communicationExisting related resourceLO2, LO6, LO7 · language and public accountabilityExisting investigationEvaluate distortion and omission in official summaries and corrections.
14. Multimodal perception and responsible decision supportExisting related resourceLO1, LO2, LO9 · perception and harmsExisting proposalTest visual and textual evidence conflicts and source reliability.
15. Evidence-calibrated social reasoningExisting related resourceLO2, LO4, LO8 · disagreement and epistemologyExisting programmeMeasure fair representation of contradictory perspectives.
16. Information science and provenanceExisting related resourceLO1, LO5, LO8 · knowledge representationExisting studyImplement historical as-of retrieval and provenance checks.
17. Education development and learning policyExisting related resourceLO2, LO6, LO7 · capability and accessExisting research programmeAnalyse education outcomes, access routes and institutional recognition.
18. Credential bottlenecks and AI assessmentExisting related resourceLO3, LO6, LO7 · governance and fairnessPublished research flagMeasure validity, reasonable adjustments and alternative competence routes.
19. Policy systems and government oversightExisting related resourceLO3, LO6, LO7 · institutions and political theoryExisting directoryCross-check governance structures, complaint routes and evidence status.
20. Legal-administrative accountability case methodExisting related resourceLO2, LO4, LO6, LO9 · contestabilityDocumented area of activity; private detail excludedUse anonymised public-law scenarios, legal sources, decision records and outcome evidence.
21. Parliamentary policy correspondence provenanceExisting related resourceLO6, LO7, LO8 · governance and communicationExisting documented correspondence categoriesSeparate sent proposal, receipt acknowledgement, substantive reply and policy adoption.
22. AV and unified communications reliabilityExisting related resourceLO1, LO2, LO9 · computing and safetyProfessional capability; cross-domain extensionTranslate commissioning, latency, availability and UAT methods into AI assurance tests.
23. Business process automation and agent servicesExisting related resourceLO1, LO3, LO9 · deployment and oversightExisting venture systems planningAssess approval gates, failure recovery, security and cost.
24. AI agent laboratoryExisting related resourceLO1, LO2, LO9 · reasoning and system designExisting demonstrationPublish controlled tool-use and evidence-retrieval test cases.
25. Research infrastructure and reproducible studiesExisting related resourceLO8, LO9 · methodologyExisting workstream portfolioUse research protocols, tests, datasets and versioned disclosures.
26. AI knowledge atlas and taxonomyExisting related resourceLO1, LO5, LO8 · definitions and historyExisting atlasAudit terms, relationships, provenance and coverage.
27. Public-good research discoveryExisting related resourceLO4, LO6, LO8 · unanswered questionsExisting initiativeBenchmark gap detection against a blinded expert review.
28. Venture opportunity and market monitoringExisting related resourceLO3, LO6, LO9 · forecasting and commercial judgementExisting monitoring pageCompare dated predictions with realised outcomes and calibration.
29. Digital tool dependency and security risksExisting related resourceLO2, LO3, LO6 · security and regulatory ethicsExisting case studiesEvaluate failure modes, contracts, data custody and dependencies.
30. Human-capability evidence and professional recognitionExisting related resourceLO4, LO7, LO8 · education and professional practiceExisting interdisciplinary research themeCompare portfolio assessment with degree screening under blinded evaluation.
31. AI-enabled compliance, transparency and audit servicesExisting related resourceLO2, LO3, LO6, LO9 · venture governanceVenture opportunity rather than operating serviceDesign documented controls, incident reporting, audit logs and independent validation.
32. Public-administration process automationExisting related resourceLO2, LO6, LO9 · public-sector accountabilityPossible application of existing architectureMeasure error correction, recourse and human authorisation.
33. Ethical data provenance and correction historyExisting related resourceLO2, LO4, LO8, LO9 · trustworthy evidenceExisting research directionDesign tamper-evident change history and assess correction latency.
34. Cross-sector risk and assurance case-study collectionExisting related resourceLO2, LO4, LO7, LO9 · case methodExisting case-study portfolioTag each study by syllabus week, method, decision type and outcome.
35. Agentic service approval and recovery testsExisting related resourceLO1, LO3, LO9 · deploymentExisting engineering roadmapTest permissions, incident response, rollback and user recourse.

Doctoral direction review: contribution and potential supervisory fit

Retain the existing fifty doctoral candidates without re-ranking or deleting them. A stronger doctoral shortlist needs an overlap audit for each candidate: (1) strongest direct prior work and date first published; (2) demonstrable unresolved limitation; (3) novel measurable contribution; (4) dataset and ethical permissions; (5) methods, baselines, falsification and reproducibility; (6) fit with a supervisor's current scholarship and university doctoral requirements; (7) feasibility, resources and expected deliverables. The research interests of lecturers and potential supervisors, LinkedIn skills/CV, decade of public posts and complete three-year conversation corpus have not been exhaustively verified here; do not claim complete alignment until their actual materials are audited.

2026 evidence: skills and qualification bottlenecks

UK government 2026 AI labour-market reporting states that 97% of surveyed respondents identified at least one AI skills gap; 35% of organisations struggled to fill AI roles. Its separate analysis of AI expert vacancy advertisements found that 99% asked for at least a bachelor's degree, with 37% requesting a PhD and 29% a master's degree. These figures describe sampled organisations and adverts, respectively, and must not be conflated with all UK employment. PwC's UK AI Jobs Barometer reports that postings requiring AI skills rose from about 112,000 to 180,000 in 2025. These findings support studying competence signals and pathways, but neither prove nor disprove unfair institutional exclusion in any particular case.

Sources: DSIT AI Labour Market Survey (released 2026); AI Skills for Life and Work; PwC UK AI Jobs Barometer (June 2026); UK cyber-skills report (2026); QAA master's characteristics.

Document and evidence integrity

This page is an additive study resource only. Neither uploaded programme specification nor teaching syllabus has been edited, deleted or replaced. Original assigned reading and optional modules remain the historical baseline. Future additions must explicitly distinguish original curricular requirements, existing documented author work, reported but unverified activities, external developments and speculative doctoral investigations. Cross-links point to the existing site without implying all areas have undergone external verification.

Interdisciplinary scientific and language atlas — 11 November 2026 expansion

Evidence status: This is an additive, source-conscious mapping across the author's submitted Levels 4–8 dossiers, AI/ML ecosystem catalogues, history syllabus, programme specification and existing Manifold Grace research. It does not replace or abridge those materials. Claims appearing in earlier AI-generated APL dossiers (including numerical 'match' estimates and statements that outcomes are met) must not be presented as independent assessor findings.

Ten disciplinary domains, their scientific boundary and research parameters

DomainConceptual breadthMethods / benchmarkLimit of inferenceExisting page
Digital mental health and social-emotional AIAffect recognition; emotion inference from text, voice and expression; empathy simulation; social intelligence; trust, therapeutic alliance, accessibility, clinical risk, consent and crisis escalationConstruct validity, subgroup performance, qualitative usability studies, prospective validation, clinician-led safety evaluationMental-state inference ≠ verified emotional understanding; clinical diagnosis requires appropriate qualified oversight.Related work
Algorithmic justice and administrative fairnessDistributive/procedural/recognitional justice, equality law, bias sources, classification thresholds, proxy discrimination, calibration, data rights, due process, explanation and remedyDisaggregated confusion matrices, subgroup calibration, legal doctrinal analysis, impact assessments, case-control comparisonNo single mathematical fairness measure settles a political or legal dispute.Related work
Human–robot interaction, HCI and embodied AIRobotics sensing, feedback, autonomy, interface affordances, anthropomorphism, assistive technologies, care robots, multimodal instruction and human factorsControlled usability and task studies, hazard analysis, simulations, failure-mode testing, HRI observationA chatbot interface, a robot and an autonomous physical actuator have distinct risk classes.Related work
Information architecture, terminology and knowledge systemsClassification, semantic models, ontologies, synonym control, provenance, taxonomy completeness, information retrieval, versioned definitions and cross-domain mappingOntology competency questions, information-retrieval metrics, expert annotation agreement, schema validationShared words can have conflicting professional definitions or jurisdictional scopes.Related work
AI engineering, data science and lab workData ingestion, preprocessing, model fitting, evaluation, APIs, monitoring, deployment, GPU/cloud/edge choices, security and costReproducible notebook, fixed dataset splits, baselines, error bars, model card, code tests and run logsA deployed interface is not by itself evidence of scientific model performance.Related work
Governance, accountability and complianceStandards, policy authority, scope of duty, audit trails, harms, responsibilities, controls, human authorisation and recourseDoctrinal research, requirements-to-control maps, assurance case, adversarial tests and field auditCompliance to one standard is not proof of ethical adequacy or absence of harm.Related work
Policy, economics and public systemsBureaucratic discretion, implementation, resource constraints, regulatory oversight, societal costs, incentives, fairness and transparencyDocument analysis, cost-benefit analysis, difference-in-differences where justified, stakeholder interviewsCorrelation and anecdotal experience do not establish institutional causal effects.Related work
AV, unified communications and reliability engineeringLatency, jitter, packet loss, PTZ control, network integrity, service performance, usability, UAT, incident response and system commissioningInstrumented tests, latency distributions, fault injection, usability tasks, defect/issue trackingReliable signal transport does not imply clinically reliable AI decisions.Related work
Business, management and enterprise transformationRisk management, process maps, operational indicators, oversight roles, automation economics, governance and controlsAction research, process mining, operational evaluation, stakeholder interviews, controlled pilotsCommercial success, academic novelty and public benefit are separately demonstrated.Related work
Education, assessment and learning recognitionCurricular outcomes, validity, academic assessment, competence, accessibility, portfolio review, supervision and doctoral originalityConstruct validation, rubric analysis, assessment comparison, inter-rater reliability, comparative institutional case studiesA self-assessed evidence portfolio is not automatically awarded credit or qualification.Related work

Separate academic vocabulary and cross-sector terminology register

Academic vocabulary covers argument, inference, theory, methods and assessment. Technical/industry terminology names specific control measures, systems and procedures. Interdisciplinary taxonomy resolves same-word/different-meaning conflicts, synonyms, hierarchical relationships, provenance, jurisdictions and changes over time. These are related but not equivalent vocabularies. The register below records 45 cross-disciplinary concepts as a versioned starting point rather than asserting a closed universal dictionary.

45 entries

#Preferred termRelated term or aliasDiscipline / arenaDefinitionResearch/application warning
1EpistemicEpistemologyPhilosophyKnowledge, warrant and grounds for beliefSpecify whether a claim concerns truth, evidence or confidence.
2OntologicalOntologyPhilosophy / information scienceWhat exists and how entities are categorisedDifferentiate philosophical ontology from a machine-readable knowledge model.
3Construct validityMeasurement validityPsychology / statistics / MLWhether a measure captures its intended constructAn emotion classifier does not necessarily measure subjective emotion.
4OperationalisationOperational definitionSocial science / analyticsTranslate a concept into measurable variablesDefine 'fairness' using a specific testable criterion.
5FalsifiabilityRefutabilityPhilosophy of scienceConditions under which a claim could be disprovedState what observation would contradict an AI-governance hypothesis.
6TriangulationCross-source corroborationQualitative / mixed methods / auditUse distinct evidence sources or methods to examine a claimAvoid treating repeated copies of one source as independent evidence.
7ReflexivityPositionalityQualitative research / ethicsExamine how the researcher and context affect inquiryKeep a bias and limitations memo.
8ProvenanceData lineage / chain of custodyInformation architecture / law / AIOrigin and transformations of an information itemTrack source, timestamp, processing and corrections.
9TaxonomyClassification schemeInformation science / biology / governanceCategories arranged by defined relationshipsMake overlapping categories and exclusion rules explicit.
10Controlled vocabularyAuthority listLibraries / data governanceStandardised permitted termsMaintain aliases and change history.
11OntologyKnowledge graph schemaComputer science / philosophyClasses, properties and relationsAvoid equating a term list with a formal ontology.
12MetadataDescriptive attributesInformation science / softwareData describing other dataRecord authorship, date, jurisdiction, version and confidence.
13InteroperabilityIntegration compatibilityHealthcare / enterprise ITExchange and meaningful use across systemsTest semantic, technical and organisational interoperability separately.
14TraceabilityRequirements-to-test mappingQA / engineering / regulationFollow a requirement through evidence and outcomeLink each learning outcome to submitted artefact and evaluation.
15SRESite reliability engineeringIT operationsReliability practices and service-level objectivesDistinguish system uptime from quality or fairness of decisions.
16SLIService level indicatorSRE / platform operationsObserved quantitative service measureE.g. request latency, error rate and availability.
17SLOService level objectiveSRE / operationsTarget for a monitored SLIDo not treat an SLO as contractual SLA.
18SLAService level agreementCommercial IT / procurementContractual service commitmentClarify remedies, measurement window and exclusions.
19ObservabilitySystem state inferabilityIT / MLOps / governanceUse logs, metrics and traces to understand behaviourAdd decision provenance, not only operational health.
20UATUser acceptance testingAV/UC / software engineeringUser validation of requirements in contextDocument acceptance criteria, defects and sign-off.
21LatencyDelayAV/UC / robotics / inferenceTime between input and observed responseMeasure physical and cognitive consequences of delay.
22JitterLatency variationNetworks / AVVariation in packet arrival timesDifferentiate from average delay or packet loss.
23RobustnessFault tolerance / distributional resilienceML / software / statisticsPerformance despite perturbationsState perturbation class and boundaries of test.
24ReliabilityRepeatability / dependabilityEngineering / psychology / measurementConsistency over time or conditionsSpecify reliability as measurement or operational property.
25ValidityInternal / external / construct validityScience / assessmentFitness of an inference or measureA reproducible result can still be invalid.
26CalibrationPrediction confidence alignmentStatistics / MLPredicted probabilities match observed frequencyTest by group and deployment context.
27Algorithmic justiceFairness / procedural justiceLaw / ethics / MLFair procedures, burdens, rights and distributive outcomesSeparate statistical parity from legitimate process and redress.
28Disparate impactIndirect discriminationLaw / organisational analyticsUnequal effects of apparently neutral decisionsCompare definition and jurisdiction before interpreting.
29ContestabilityChallenge / appeal / redressAdministrative law / AI governanceAbility to question and correct decisionsTest access to reasons, human review and effective remedies.
30Affect recognitionEmotion inferenceAffective computing / psychologyInfer affect-related states from signalsDo not equate facial/voice signals with direct access to feelings.
31Empathy simulationConversational empathyHCI / digital mental healthBehaviour that appears understanding or supportiveDistinguish perceived empathy from actual clinical competence.
32Theory of mindMental-state attributionPsychology / AI evaluationInference about beliefs, intentions and perspectivesDistinguish task success from a real mental model.
33Social intelligenceSocial cognitionPsychology / conversational AIUnderstanding interactional norms and contextsMeasure context sensitivity, fairness and repair of misunderstandings.
34Human–robot interactionHRI / human–machine interactionRobotics / HCI / safetyHuman interaction with embodied automationEvaluate trust, physical safety, autonomy and accessibility.
35Human-in-the-loopMeaningful human oversightMLOps / governance / lawHuman review within automated workflowAssess actual authority, time and ability to override.
36Critical realismRealist methodologySocial science / philosophy of scienceMechanisms and structures beyond surface observationsDistinguish observed events from inferred causal mechanisms.
37Mixed methodsMethodological integrationInterdisciplinary researchCombine qualitative and quantitative evidenceJustify integration rather than presenting two unrelated analyses.
38Case studyBounded empirical investigationSocial science / law / engineeringExamine a defined case in contextSpecify case boundary, evidence provenance and generalisation limits.
39Grounded theoryInductive theory generationQualitative researchDerive conceptual categories through iterative codingDocument coding, theoretical sampling and saturation rationale.
40Difference-in-differencesQuasi-experimental designEconometrics / policyCompare outcome changes across groupsCheck parallel trends and interference assumptions.
41LongPESTLeExtended strategic environmental scanStrategy / policyPolitical, economic, social, technological, legal, environmental, ethical/spiritual factorsDefine each factor, evidence and interactions; not a substitute for causal inference.
42Scientific integrityResearch integrity / TEVVScience / AI assuranceTransparent and defensible design, testing and evaluationRegister methods, sources, uncertainty and reproducibility.
43Evidence thresholdBurden / standard of proofLaw / research / quality assuranceMinimum support necessary for a claimDo not import legal standards into statistical inference without definition.
44Semantic interoperabilityMeaning preservationHealthcare / information architectureSame meaning retained across systemsTest mappings, synonyms, scope and version differences.
45RPL / APLRecognition of prior learningHigher education / accreditationInstitutional evaluation of relevant prior learningSeparate portfolio claim, assessor decision and formally awarded credit.

The vocabulary should not be reduced to memorisation: examine when a word represents a different construct in two fields (for example, reliability in psychometrics versus operations; validity in assessment versus system testing; ontology in metaphysics versus knowledge representation).

Research methodologies — discipline-dependent best practice and boundaries

ArenaPossible established methodsMinimum parameters to define before use
Computing/AIDesign science, experiments, benchmark studies, formal verification, software testingTest set provenance; baseline; error rate; reproducibility; robustness; compute cost
StatisticsExperimental/quasi-experimental designs, Bayesian/frequentist modelling, causal inferenceSampling, assumptions, uncertainty, effect size, calibration, sensitivity
Psychology/digital mental healthValidated scales, observational methods, interviews, trials where authorised, psychometricsConstruct validity, consent, safeguarding, sample diversity, confounding
Human factors/roboticsUsability testing, task experiments, contextual inquiry, simulation, hazard analysisTask success, workload, safety, operator override, accessibility
Law/governanceDoctrinal analysis, comparative law, case analysis, regulatory impact assessmentAuthority, jurisdiction, interpretation, precedents, procedural remedy
Policy/economicsProgramme evaluation, econometrics, institutional comparison, process tracingCounterfactuals, baseline, cost, distribution, assumptions, unintended effects
Management/operationsCase studies, action research, operational performance evaluation, process miningProcess metrics, decision rights, stakeholder reliability, change attribution
Information scienceOntology engineering, retrieval evaluation, content analysis, classification studiesInter-annotator agreement, precision/recall, semantic coherence, provenance
Media/cultureContent analysis, discourse analysis, historical methods, audience researchCorpus design, coding protocol, contextual interpretation, reflexivity
Interdisciplinary scienceMixed methods, triangulation, cross-case synthesis, systems modellingShared units of analysis, coherence of integration, error propagation

For every proposed research method record question, unit of analysis, time horizon, population, inclusion/exclusion, sampling, ethical permissions, comparator, data provenance, collection instrument, reliability, validity, analysis plan, limitations and disclosure of prior work. No method is automatically best without checking the research purpose and context.

What qualifies as scientific, technological, engineering or mathematical evidence?

Kind of inquiryRigour conditionsBoundary / caution
Scientific inquiryTestable or otherwise systematically evaluable claims; transparent methods; observable or analytically justified evidence; uncertainty; reproducible or auditable reasoning; critical reviewExperimental, observational, modelling and some qualitative sciences can meet these criteria.
Scholarship and humanitiesInterpretive arguments, conceptual analysis, historical and legal interpretation, text and archival criticismRigour does not require turning every philosophical or legal question into an experiment.
Technology developmentWorking system, documented requirements, usability, safety and performance testingA functioning artefact is engineering evidence; not necessarily a novel scientific finding.
EngineeringDesign constraints, specifications, validation and verification, hazard control, repeatable measurement and traceabilityExplicitly distinguish verification ('built right') from validation ('right problem').
MathematicsDefinitions, formal assumptions, proofs, derivations and counterexamplesMathematical proof differs from empirical support and policy argument.
InterdisciplinarityAppropriately integrated concepts, compatible constructs, cross-field methods and a coherent research questionBreadth alone is not demonstrated synthesis; integration should yield testable, defensible insight.

LongPESTLe and checks against omitted dimensions

The uploaded Comprehensive AI/ML Ecosystem Dossier explicitly states the LongPESTLe dimensions as Political, Economic, Social, Technological, Legal, Environmental and Ethical/Spiritual. Preserve those seven dimensions as the author's recorded overlay. To check coverage, cross-reference systems engineering, scientific validity, institutional powers, privacy/security, accessibility, culture, time, geography, implementation, maintenance, accountability, supply chains, distribution of effects and remedies. These are additional analytic checks, not claims that the uploaded acronym formally contains more letters or that every proposed extension is original.

For each factor: identify its operational definition; affected groups and jurisdiction; date of observation; supporting data; uncertainty; direct and indirect effects; conflicts with other factors; and why its omission would affect conclusions. The MILES research programme, completeness and stewardship study and framework coverage atlas are existing linked research, not automatically verified empirical achievements.

Lab sessions — demonstrable learning-outcome links

Lab exercise or existing resourceDirect educational purposeEvidence suitable for reviewLink
AI research and data evaluationLO1, LO2, LO8, LO9: model operation, evidence and research designNotebooks, datasets, versioned run records, error analysesExisting related page
Interactive AI systems and agentsLO1, LO3, LO9: architecture, human oversight, evaluationSpecifications, test suites, screenshots, source commits, risk logsExisting related page
Reasoning and retrieval experimentsLO2, LO4, LO8, LO9: evidence comparisons and falsificationGround-truth register, cases, baseline comparisonsExisting related page
Bias and reliability studyLO2, LO4, LO6, LO9: fairness and limitationsSubgroup metrics, measurement validity and findingsExisting related page
Clinical/social-emotional simulation — proposed research applicationLO2, LO4, LO6: therapeutic risk and emotional inference limitsSimulated non-patient tests, ethics protocol, professional reviewExisting related page
Robotics/HRI simulation — proposed research applicationLO1, LO2, LO9: action, feedback, accessibility, safetyTask-level simulation, hazard controls, override testsExisting related page
AV/UC commissioning case study — professional evidenceLO1, LO9: control systems, signal quality, UAT and troubleshootingRedacted signal measurements, decision logs, test protocolsExisting related page
Government/policy evidence validationLO3, LO6, LO7, LO8: institutions, interpretation and responsibilityAnonymised records, legal source map and correctionsExisting related page

These links do not establish that each hypothetical experiment has been completed. Assess each real session against its actual run date, code, source, outcome and independence of supervision.

Evidence crosswalk: uploaded Levels 4–8 and discipline breadth

LevelSubjects / dossier categoriesDocument sourceReported evidence typesWhat a fair reviewer must distinguish
4Psychological sciences: development, observation, safeguarding, research methods; computing: web, networks, security, UX, databases and digital literacyLevel 4 APL dossier (uploaded)Early years and applied professional examples; CIW and digital skillsClaims marked ‘met’ are dossier assertions; require module-specific evidence and authorised review.
5Web design, coding, network architecture, HCI/accessibility, information security, digital business and marketingLevel 5 APL dossier (uploaded)CIW foundations, site development, networking, JavaScript and user-facing practicePrior percentages are draft alignment estimates, not institutional RPL determinations.
6Advanced web engineering, networks/security, data modelling, HCI, digital production, business/governance, LawTech and AI-adjacent computingLevel 6 APL dossier (uploaded)Systems integration, PTZ/AV, QA/UAT, SOPs, accessibility and complianceDistinguish direct artefacts from claimed transferability to specific modules.
7Postgraduate web/HCI/digital media, selected computing governance, business, law/new technologies, PPE, education and sustainabilityLevel 7 APL dossier (uploaded); MA AI Ethics & Society dossierSynthesis, evaluation, independent reasoning and authored research/policy documentsAssess against exact outcomes and awarding rules, not generic 'postgraduate match' claims.
8Research directions in management, computing/HCI, digital media, occupational psychology, education, law and interdisciplinary AI ethicsLevel 8 APL dossier (uploaded)Potential starting questions, concept frameworks, practical case studies and doctoral proposal evidenceThe dossier itself explicitly does not establish a completed doctorate; originality needs rigorous prior-art and thesis examination.

Source-integrity note: Levels 4–8 PDF dossiers and the three AI/ML ecosystem documents were examined as user-supplied mapping proposals. They are not awarding-body certificates and cannot prove APL or doctoral equivalence by themselves. The Level 8 document itself explicitly says it does not claim completion of a doctorate. Similarly, the 2025 MA feedback dossier documents the author's concern that disciplinary analytic writing and broader interdisciplinary synthesis should be assessed using clearly specified criteria; it is a recorded submission, not an independent finding of maladministration.

Beyond the 2024–25 AI syllabus and programme specification

Extensions, deliberately separated from formal MA requirements: digital mental health and emotion inference, advanced robotics/HRI, AV and UC systems, reliability engineering, detailed machine-learning platforms/MLOps, sector terminology, formal taxonomy/ontology engineering, operational management and procurement, legal advocacy, accreditation and RPL, longitudinal governance audit, integrated framework-completeness tests, workplace QA and business-process automation. These expand research possibilities but do not retrospectively redefine the course or constitute universally certified expertise.

Sources, institutional standards and verification

NIST AI Risk Management Framework (govern/map/measure/manage, AI system trustworthiness) · ISO 9241-210:2019 (human-centred design; confirmed 2025) · ISO/TR 23482-2:2019 (safety guidelines for close human-robot interaction) · WHO 2025 AI health guidance · WHO 2021 AI health ethics · Official England qualification levels · CESI evidence guidance.

For institutional review, every asserted learning outcome should be supported by a particular artefact, date, provenance, relevant module descriptor, applied method and independently checkable result. This improves clarity and reduces ambiguity; it cannot prevent legitimate disagreement or compel an institution to award credit.

Applied learning evidence: AI authorship, intellectual property and non-bypassable system governance

11 November 2026 · additive research evidence statement. This section records a sustained author-reported line of investigation and documented research discussions, rather than merely familiarity with textbook concepts. It distinguishes a person's initial questions, domain experience, corrections and intellectual direction from assistance provided by AI systems in presenting, comparing and refining them. Specific original dates and any claims about who first proposed a mechanism require dated primary-source verification; the publication date of this section is not its conception date.

Contribution provenance: corrections, authorship and responsible AI assistance

Research or educational activitySubstantive question and applicationEvidence and appropriate qualificationProgramme and doctoral relevance
Correcting AI responses on intellectual property and creditChallenges to misleading attribution, conflation of source information with authorship, and unsupported claims that an AI assistant independently originated the user's conceptual proposals.Author reports multiple corrections and discussions over preceding months; verify verbatim cases, dated transcripts, original prompt and response, correction accepted or rejected, and any later system response. A conversation correction does not by itself demonstrate a permanent model change or legal entitlement.Ethics, epistemology, intellectual property, evidence standards, human–AI collaboration, critical evaluation and researcher provenance.
AI personhood and the allocation of creditSeparates functional agency and apparent intelligence from legal personhood, moral status, inventor status and human responsibility. An AI system is not presently a human legal author merely because it generated text.Documented research theme and author-reported intellectual position. The legal implications vary by jurisdiction and doctrine; copyrightability depends on sufficient human authorship and is not determined by who typed the most words.Philosophy of mind, philosophy of law, responsibility, accountability and contested notions of agency.
AI-assisted education and original intellectual directionThe user supplies questions, cases, arguments, corrections, hypotheses and practical experience; search engines, readings, AI dialogue and automated structuring support inquiry. The use of an assistant is a contribution to learning, not proof that the assistant conceived the questions or that all output is solely human-authored.Evaluate original dated prompts, successive revisions, independently written work, outputs tested against scholarly sources and disclosures of tool assistance. Record where AI materially proposed a concept as well as where it only edited or organised one.Independent inquiry, metacognition, scholarly integrity, formative feedback, information literacy and doctoral methods.
Real-world organisational application and social consequencesTransfers concepts from books and lectures into enterprise systems, cybersecurity, public administration, rights, regulatory oversight, workplace design, unequal access and downstream harms.Use verifiable professional artefacts, anonymised case records, policy submissions, software analyses and independent source checks. General discussion of a scenario is not proof of a deployed intervention.Applied socio-technical research, case methodology, risk-based policy analysis and research impact.

Proposed guardrails and cross-checking agents

The following are research and engineering design proposals, not claims that a regulator has certified an implementation. No software-only safeguard can be assumed literally impossible to bypass in every threat model; the stronger aim is separation of control, least privilege, independent verification, tamper-evident records and human or institutional authority outside the model's power.

Proposed control or assurance agentPurpose and boundaryEvidence/test required
External fixed policy control planeStore and enforce critical permissions, policy configurations, compliance gates and supervisory approvals outside a self-editing AI agent; restrict model authority to change them.Threat model, independent access-control test, least-privilege review, attempted unauthorised-change tests and recovery procedures.
Architecture and configuration attestationIndependently compare approved architecture, deployed versions, dependencies, agent tool access and changes to detect concealed or unauthorised modifications.Signed build artefacts, deployment attestations, reproducible checks, configuration drift detection and external audit.
Action-transparency and user-notification agentProvide intelligible descriptions of what an AI system is attempting, which tools it used, what data categories were accessed, which constraints apply, and where uncertainty remains.Compare action claims with independent event logs; measure omissions, misleading explanations and usability. Textual self-report alone is not a trustworthy audit source.
Provenance and intellectual-contribution verifierDistinguish user-authored content, retrieved material, AI-generated additions, revisions, and linked sources, including whether a proposed idea appeared in earlier records.Versioned drafts, hash-linked timestamps, source-lineage review, contradiction audit and human adjudication. A timestamp proves existence by a date, not global originality.
Regulatory requirement mapping agentMap system function, jurisdiction, risk category, accountable organisation and operational controls against applicable statutory obligations and standards.Versioned legal source register, human legal review, jurisdictional scope and dated regulatory changes. Such mapping cannot itself determine legal compliance conclusively.
Independent adversarial/cybersecurity assuranceCheck prompt injection, tool misuse, data exfiltration, model/configuration tampering, security boundaries, logging integrity, rollback and escalation.Authorised security testing, reproducible fault and misuse cases, separation of duties, detection/response metrics and independent sign-off.
Social-impact and contestability reviewTest potential harms to privacy, employment, education, attribution, accessibility, due process and affected third parties; ensure appeal and correction paths.Distributional evaluation, stakeholder evidence, accessible challenge mechanisms, remedies and post-deployment review.

Governability problem: innovation may outpace effective oversight

A proliferation of AI enterprises, autonomous agents and opaque changing architectures could create verification and enforcement gaps even where relevant law applies. The hypothesis to test is not that these companies are inherently unregulated, but whether complex outsourcing, rapid version changes, opaque model dependencies and mutable agent control surfaces prevent regulators and purchasers from identifying duty holders, establishing evidence and enforcing obligations. Analyse business models, architecture changes, audit access, standards, cyber resilience, remedies and jurisdictional reach separately.

Existing legal and standards context (not original inventions)

The EU AI Act includes high-risk system provisions concerning automatic logging (Article 12), human oversight (Article 14) and cybersecurity/robustness (Article 15). These are related prior standards, not proof that any independent proposal preceded them, and applicability depends on the system and current legal implementation. EU Article 12 · EU Article 14 · EU Article 15. NIST Generative AI Profile (2024) supplies an established voluntary governance framework. US Copyright Office report (2025) explains that use of AI assistance does not automatically disqualify human-authored work but that insufficiently human-directed purely AI-generated expression is not protected under US copyright law. Do not generalise this position unchanged to UK law or patent inventorship.

Learning-objective crosswalk and falsifiable doctoral extensions

Learning dimensionEvidence already identified or reportedPotential formal research testSource / connected work
Philosophy of mind, agency, authorshipResearch discussions distinguishing AI processing, legal personhood, moral responsibility and rights of human contributors.Compare candidate attribution taxonomies against law, philosophy and blind reviewer assessment; test inter-rater agreement.AI auditing history · Information science
Ethics and social impactsAuthor-reported corrections and enquiries into misleading credit, workplace accountability, errors and institutional consequences.Audit provenance and correction outcomes under human–AI authorship tasks with documented source histories.Evidence verification · Social reasoning
Computing and system designConceptual proposals for independently enforced controls, architecture-change transparency and multi-agent checks.Prototype and compare model-only rules against an external control plane and independently monitored agents; measure unauthorised-change detection and false alarms.Agent assurance · Implementation roadmap
Cybersecurity and operational reliabilityCross-sector professional reasoning in IT, networked AV/UC, validation, risk and governance.Perform authorised threat-model-based evaluations of access boundaries, update integrity, audit traces and recovery.System risk cases · Venture systems
Law, policy and regulatory scienceProposed corporate safeguards and monitoring agents, connected with legal advocacy, public accountability and policy correspondence.Construct a jurisdiction-specific obligations-to-controls matrix and independently score traceability and regulatory explainability.Framework completeness · Public systems
Research methodology and independent learningQuestions, dated conversation traces, revisions and document synthesis can demonstrate iterative inquiry when securely evidenced.Chronologically reconstruct proposal origin, strongest prior art, revisions, independent tests and counterexamples; disclose material AI assistance.CESI evidence guide · Research methods

Reviewer standard: For each asserted contribution record the exact claim, date first evidenced, primary artefact, AI contribution and human contribution, closest earlier scholarship, application, independently verifiable result and any unresolved uncertainty. This demonstrates intellectually directed learning beyond passive consumption without asserting that every discussion is a novel discovery, legally owned invention or independently confirmed competence.