University of Jyväskylä Artificial Intelligence Master's Thesis Guide: AIAS5002, Thesis Types, IT Assessment and Vasara

Start with the current Artificial Intelligence thesis package

The current University of Jyväskylä Artificial Intelligence master’s degree uses a 36 ECTS compulsory thesis-studies block: TIES5011 Research Methods in Computer Science, 3 ECTS; AIAS5001 Master Thesis Seminar for IMDP AI students, 3 ECTS; AIAS5002 Master’s Thesis, 30 ECTS; and AIAS5003 Maturity Test, 0 ECTS. TIES5011 and AIAS5001 are Pass/Fail, AIAS5002 is graded 0-5, and AIAS5003 is Pass/Fail. These are the programme-specific current objects under AINMP2024 / AIADV for the 2024-2028 curriculum.

1. The current degree is Master of Science

Artificial Intelligence is a 120 ECTS, two-year, English-language Master of Science programme in the Faculty of Information Technology. The current Study Guide route is AINMP2024, with curriculum periods running from 2024-2025 through 2027-2028. The identifier does not mean the curriculum ended in 2024. Use the current Study Guide and course objects rather than older programme material when planning thesis registrations.

2. AIADV is the current advanced-studies module

The current Advanced Studies in Artificial Intelligence module is AIADV, with a scope of 96+ ECTS. The programme develops competence across data-driven AI, knowledge-based AI, and autonomous and responsible AI. Those pillars help define the disciplinary environment, but they do not force one thesis architecture. A thesis can address an AI problem through empirical research, systematic literature work, constructive development or an article-based route where the Faculty conditions are met.

3. The compulsory thesis studies total 36 ECTS

The exact thesis-study block is easy to misstate because the degree contains other project and writing-support courses. The controlling compulsory thesis studies are TIES5011 3 ECTS, AIAS5001 3 ECTS, AIAS5002 30 ECTS and AIAS5003 0 ECTS. XENB0007 Master’s Thesis Support, 2 ECTS, belongs to communication and language studies rather than this 36 ECTS block. Keep these components separate in Sisu and in any description of the programme.

4. TIES5011 builds research-method literacy before the thesis

TIES5011 Research Methods in Computer Science, 3 ECTS, Pass/Fail covers central areas of computer-science research methodology, foundations of scientific activity, the role of methods in a research process and ethically sustainable research. It prepares students for thesis work through method literature, reading-circle activity and evaluation of completed MSc theses. The course is not a checklist of one mandatory AI method. Its purpose is to help students recognise and judge methods appropriate to different research questions.

5. TIES5011 expects sustained participation

The current course implementation includes six reading circles, contributions to their reporting and six individual evaluations of MSc theses using the methods discussed. It therefore requires regular weekly work rather than a single final assignment. A BSc thesis is listed as a prerequisite. Use the course to compare how different studies justify evidence, design, analysis and validity, because the IT Faculty later assesses these dimensions explicitly in the master’s thesis.

6. AIAS5001 is the programme-specific thesis seminar

AIAS5001 Master Thesis Seminar for IMDP AI students, 3 ECTS, Pass/Fail supports the final stage of the degree. It covers topic selection, literature search and analysis, selection of research methods, development of a research plan and presentation of thesis components. Before the seminar, the topic is expected to be agreed with a thesis supervisor and approved by the Head of the study programme. The seminar is designed to turn a broad AI interest into an executable thesis plan.

7. Read the seminar’s empirical-part wording carefully

The current AIAS5001 description includes design and implementation of an empirical part. That wording is important for the seminar, but it must not be turned into the false claim that every acceptable AIAS5002 thesis must be empirical. The Faculty of Information Technology separately permits empirical, literature-review, constructive or developmental, and article-based master’s theses. The final thesis form must follow the Faculty rules, the approved topic and the supervisor-agreed research or development design.

8. AIAS5001 has readiness expectations

The seminar expects a bachelor degree, an up-to-date study plan compatible with graduation within one year of starting the thesis, and at least preliminary consideration of the topic and supervisor. For IMDP AI students, TIES4570 Cognitive Service Development Project is a prerequisite. Current completion also requires active seminar participation, weekly homework and a final presentation. Treat the seminar as a working thesis environment, not as a replacement for AIAS5002.

9. AIAS5002 is the controlling 30 ECTS master’s thesis

AIAS5002 Master’s Thesis, 30 ECTS, grade 0-5, is the programme-specific thesis object. It is completed in English as independent study. The IT Faculty describes the master’s thesis as a written academic work and equates 30 ECTS with approximately six months of full-time study. The thesis must demonstrate the ability to define and justify a research problem or development task and carry the work through as a coherent scientific project.

10. The thesis needs scientific grounding even when it develops technology

Current Faculty learning objectives require students to obtain, interpret, synthesise and critically analyse scientific sources; understand previous research and relevant theories; select and apply a suitable research or development methodology; evaluate methodological choices critically; follow research ethics and good scientific practice; relate results to prior knowledge; report scientifically; and manage the overall research or development process. A working system, model or application alone is therefore not sufficient.

11. Empirical research is one valid thesis route

An empirical research-based thesis uses empirical data and relevant scientific research methods to produce well-founded results connected to previous research and theory. In AI, this might involve experiments, benchmark data, simulations, user studies, system measurements or other evidence, depending on the question. The current sources do not require every empirical thesis to train a new model, use deep learning or beat a benchmark. Method choice must be justified by the research problem.

12. A literature-review thesis is also possible

The IT Faculty permits a literature-review thesis, but it is not a simple narrative summary of papers. The review must be systematic and comprehensive and use an accepted literature-review or mapping methodology. The Faculty also expects a significant original contribution, such as a new model, framework or other synthesis with novelty value. Plan the search strategy, inclusion criteria, analytical framework and contribution explicitly so that the work is assessable as research rather than as reading notes.

13. Constructive or developmental theses are permitted

A constructive or developmental thesis may create an artefact such as software, an algorithm, a model, a method or material. The artefact must be linked to prior research, and its effectiveness or functionality must be evaluated. This is particularly relevant to AI students, because building a technical system can feel like the whole project. The Faculty framework requires the development task, scholarly context, methodology and evaluation to form one thesis, not a product plus a loosely attached report.

14. Article-based theses have additional conditions

The IT Faculty also permits an article-based master’s thesis where the stated publication and workload conditions are met. The route includes one or more scientific articles meeting the Faculty threshold plus a literature-review introduction. The student’s independent and significant contribution must be identifiable, and the total workload must correspond to the normal thesis ECTS scope. Do not assume that co-authorship on a paper automatically satisfies AIAS5002 without checking the current Faculty conditions.

15. Pair and dual-degree arrangements are special cases

The Faculty guidance also recognises jointly written theses and dual-degree arrangements under specific rules. These are not the default route. If your project is shared with another student or another institution, establish authorship, independent contribution, supervision and assessment arrangements before substantial work begins. The Degree Regulations still require the student contribution to be assessable and the thesis process to remain compatible with JYU’s formal evaluation requirements.

16. TIES4570 Cognitive Service Development Project is not the thesis

TIES4570 Cognitive Service Development Project, 15 ECTS, grade 0-5, is a compulsory advanced-studies project in which teams develop a cognitive service from a business idea through design and implementation to cloud launch. It includes teamwork, testing, service audit, a report, public defense, a business model and IPR arrangements. It may generate thesis topics and is a prerequisite for AIAS5001, but it is not AIAS5002 and does not by itself satisfy the 30 ECTS thesis requirement.

17. Project work can feed a thesis only through a new scholarly design

A useful TIES4570 result can become background, a research artefact, a case or an empirical platform for a thesis when this is agreed with the supervisor and the thesis has its own justified research or development task. Reusing code or a project idea does not remove the need for literature, methodology, evidence, evaluation and scientific reporting. Define what is genuinely new in the thesis and what came from earlier group work so the individual contribution is clear.

18. Deep learning is compulsory coursework, not a universal thesis method

The programme includes compulsory deep-learning and cognitive-computing content, including TIES4910, which covers several learning paradigms. This disciplinary depth does not mean every thesis must use a deep neural network. An AI thesis may investigate symbolic reasoning, knowledge representation, agent systems, system evaluation, responsible AI, literature synthesis, software engineering around AI, or another approved problem. Choose the method because it answers the thesis question, not because a fashionable model family appears in the curriculum.

19. Knowledge graphs and agents are also optional at thesis level

Current compulsory studies include semantic-web and linked-data content and collective-intelligence or agent-technology content. That prepares students to work across multiple AI paradigms, but it does not require every AIAS5002 thesis to use RDF, ontologies, knowledge graphs, autonomous agents or multi-agent systems. The same principle applies to programming languages, frameworks and cloud platforms: the current registered sources do not impose one universal technology stack for the thesis.

20. Find the responsible supervisor early

The IT Faculty thesis process begins when a responsible supervisor is appointed and approves the topic. The responsible supervisor must hold a doctoral degree and be employed by the University of Jyväskylä. A thesis may have additional supervisors, including people with specialised expertise, but the responsible-supervisor rule still controls the formal process. AI students are advised to seek a supervisor well before the intended start of the thesis rather than waiting for the seminar to solve topic and supervision questions automatically.

21. There is no centralised AI thesis-topic list

Current AI guidance says there is no single central list of Artificial Intelligence thesis topics. Students can identify possible supervisors by looking at faculty teaching, research groups and research profiles and by networking with instructors. Research groups may occasionally advertise current topics, and study advisors can help. Students generally have meaningful freedom in selecting a topic and possible commissioning entity, so a company partnership is an option rather than a programme-wide thesis requirement.

22. Agree the supervision process, not only the topic

Once the responsible supervisor and topic are established, agree how the work will proceed: the working research question or development task, thesis form, expected evidence, milestones, feedback rhythm, data and ethics responsibilities, and what constitutes a draft ready for formal submission. The Faculty describes the thesis as approximately six months of full-time work, but calendar duration can vary with the research design, access to data, employment and other studies. A clear supervision agreement helps keep that workload bounded.

23. The IT Faculty grading framework is cumulative

The IT Faculty evaluates the thesis across multiple criteria, including the research problem, familiarity with the topic and earlier research, execution and method, critical assessment, interpretation and reflection, scientific reporting, and the overall research or development process. The criteria are cumulative: stronger grades require the qualities expected at lower levels plus additional quality. Every evaluated criterion must reach at least the satisfactory grade-1 level for the thesis to be accepted.

24. The final grade is not an arithmetic average

The examiners’ proposed overall grade is not calculated as a simple mean of criterion scores. The Faculty guidance states that at least half of the evaluated criteria must meet the level proposed for the overall grade. Intermediate grades 2 and 4 are used even though they are not separately described in the published matrix. The Vice Dean determines the final grade on the basis of the examiner assessment under the Faculty process.

25. Strong technical performance cannot compensate for a missing research foundation

A technically impressive model or system does not automatically produce a high thesis grade. The IT criteria look at whether the problem is meaningful and justified, whether previous research and concepts are handled comprehensively, whether the method matches the objectives, whether data or evidence are handled carefully, whether results are interpreted critically and whether the scientific report is coherent. For grade 5, the matrix expects particularly strong justification, command of literature, method-evidence alignment and analytical discussion.

26. AIAS5003 is the programme-specific maturity object

AIAS5003 Maturity Test (MSc), 0 ECTS, Pass/Fail, is the current Artificial Intelligence maturity-test course. A maturity examination is required before graduation and demonstrates familiarity with the thesis topic together with applicable language competence. The general JYU maturity instructions determine the language route based on the student’s educational background and earlier demonstrations. Use AIAS5003 rather than copying the maturity code from another JYU programme.

27. Human-user data can create personal-data obligations

AI research may involve interviews, usability studies, logs, platform identifiers, behavioural traces, images, audio, location-derived features or other information relating to people. JYU treats information about an identified or identifiable living person as personal data. Online identifiers and combinations of contextual information can also identify a person. If you independently conduct graduation research for your own purposes, you are normally the data controller and must plan the processing before collection.

28. Pseudonymised data can still be personal data

Replacing direct identifiers with codes does not automatically make a dataset anonymous. If a participant can still be re-identified through a key, rare attributes, timestamps or linked records, the material can remain personal data. This matters for AI projects because high-dimensional datasets can carry identification risk even after obvious identifiers are removed. Describe realistic access controls, retention, deletion, participant information and re-identification risk rather than using “anonymous” as a default label.

29. Ethical review is conditional, not automatic

JYU human-sciences ethical review follows TENK principles. Every project must be conducted ethically, but formal prior committee review is required only when the applicable criteria are met. A user study does not automatically require committee review merely because humans participate. Conversely, projects involving significant intervention, vulnerable groups, sensitive contexts or other TENK triggers should be checked early. When prior review is required, obtain it before the relevant data collection or research begins.

30. Use a Data Management Plan for datasets, code and derived outputs

JYU’s Data Management Plan framework covers collection, use, storage, reuse and disposal of research data, as well as legal and ethical issues, rights, responsibilities, documentation, metadata and security. For an AI thesis, the plan can include raw data, labels, preprocessing outputs, model checkpoints, code, configuration files, evaluation results and documentation. Clarify which materials can be shared, which are restricted, and how the thesis can remain reproducible within legal, ethical and contractual limits.

31. Using AI tools in an AI thesis still follows JYU’s AI rules

Studying artificial intelligence does not create an exception to the University’s rules for using AI-based applications in assessed work. Relevant AI assistance must be reported transparently, AI output is not a scientific source, and the student remains responsible for factual accuracy, analysis, citations and the final text. Claims suggested by an AI system must be checked against original sources. Do not upload personal, confidential or restricted research material to an unapproved service.

32. Citation and accessibility remain formal thesis obligations

JYU requires sources to be indicated when paraphrasing another person’s ideas and when quoting directly, and the selected citation style must be used consistently. The current registered AI sources do not establish one universal named style edition such as APA for every thesis. The final online thesis must also meet accessibility requirements. Prepare the document structure, figures, tables, alternative text and final PDF/A workflow early rather than treating accessibility as a last-minute conversion problem.

33. Vasara submission means the thesis is final

JYU uses Vasara for electronic submission, evaluation and archiving of master’s theses, and the Faculty of Information Technology directs Artificial Intelligence students to the general return-for-review process. The file submitted to Vasara must be the final version intended for assessment. The applicable Turnitin check must be completed before submission, and confidential background material must be removed from the public thesis. After formal submission, ordinary editing of the assessed work is no longer available.

34. Two examiners and response rights protect the assessment process

Under JYU Degree Regulations, two examiners are nominated for the master’s thesis, at least one of whom must hold a doctoral degree; the second examiner may be the supervisor. The examiners provide an evaluation statement and grade proposal, normally within one month of final submission. The student has an opportunity to respond to the examiner statements before the final grade and also has the applicable right to interrupt assessment before the final decision.

35. The approved thesis becomes a public academic work

The evaluated master’s thesis is sent to the JYX publication archive under the current workflow, and master’s theses are public academic works. Plan confidentiality accordingly. If a company, research group or external data provider supplies restricted material, the public thesis still needs enough methods, evidence and reasoning to be academically assessable without exposing protected information. Once an accepted thesis is graded, it cannot simply be rewritten or retaken to raise the grade.

36. Build the thesis timeline around topic, method, seminar and final review

A reliable sequence is to use TIES5011 to strengthen method literacy, complete the compulsory AI studies and TIES4570, identify a feasible topic and qualified responsible supervisor early, and enter AIAS5001 with the study plan and thesis direction already taking shape. Use the seminar to develop literature, method and research plan, then reserve the main AIAS5002 period for execution, analysis, evaluation and writing. Leave time for supervisor review, Turnitin, maturity requirements, accessibility checks and final Vasara submission.

37. Final Artificial Intelligence checklist

Verify that Sisu reflects the current AINMP2024 / AIADV route and the exact 36 ECTS thesis-studies package. Confirm the topic and responsible JYU-employed doctoral supervisor, choose the thesis type explicitly, and justify the research or development method rather than defaulting to deep learning or a preferred tool stack. Keep TIES4570 separate from AIAS5002, plan personal-data and ethics obligations before collection, maintain a Data Management Plan, document AI-tool use, evaluate the work against the cumulative IT Faculty criteria, complete AIAS5003, Turnitin and accessibility requirements, and submit only the final assessment version through Vasara for eventual JYX publication.