Build an early-career research portfolio around supervised work

1. Turn the ambition into something a mentor can inspect

Start with a capability expressed as a task: distinguish a research question from a policy ambition, check a source against a claim, or explain why a result does not support a causal conclusion. Then name the artifact that would show the task was attempted. A source-check table is easier to review than a statement that the learner has “mastered evidence”. The mentor should be able to point to both the original attempt and the correction.

Research assessment can recognize outputs beyond published papers. DORA’s recommendations encourage attention to the content of research and to a range of contributions, including data and software. Here that informs a proposed portfolio approach; it does not establish a Saudi institution’s promotion policy or promise formal credit. Agree locally which work can be assessed and by whom, without making acceptance by a journal the only visible endpoint.

Reference: DORA: assess research on its merits and recognize diverse outputs

2. Choose a small exercise with a real supervision boundary

For a first cycle, propose a synthetic exercise that can be inspected without participant records. The learner could refine a service question, compare two primary sources, construct a tiny invented data dictionary, prepare a script and explain its eventual output. A mentor identifies the point at which each artifact needs review. Agree which materials may be used and where the work is stored; a practice task should not quietly become an unauthorized extract from a service database.

Use a progression that follows dependencies. A script written against an unresolved unit of measurement cannot be treated as an analysis result. A polished discussion cannot establish what an unrun script would show. Ask the learner to identify the next missing input and propose a limited way to resolve it. This is a suggested sequence to adapt to the team and discipline, not a validated training package or a prediction of learning gains.

On small screens, scroll the table horizontally, or focus it and use the arrow keys.

3. Keep training separate from authorization for a study

The Saudi National Committee of Bioethics researcher portal describes ethics training and a route for submitting research proposals for review. Confirm the current training requirements and certificate validity with the relevant committee. Completing a training course is different from receiving a decision on a particular project, version, site and activity. This guide neither determines the required review route nor grants access to data.

Make that distinction practical in supervision. Label the synthetic dataset as invented. If a future task needs real material, ask the appropriate institutional office to clarify the applicable permissions before that activity starts. Record the decision and its scope rather than assuming a supervisor’s enthusiasm, a seminar attendance record or a course certificate covers it. A learner should be able to explain what they are currently allowed to do and what remains a proposal.

Reference: Saudi NCBE: researcher training and proposal review

4. Review the portfolio against its actual state

Consider a wholly fictional learner record. Six artifacts are planned. The question note has been corrected and the source comparison checked by the mentor. The dictionary still has an unresolved unit, the script has not been run, no results are available for interpretation and the full review note remains pending. Only the first two artifacts have a documented completed review in this example. That is not evidence of completing a course or of being ready to conduct an independent study.

The immediate issue is the dictionary’s unresolved unit. If a draft variable is called “duration”, the learner must establish whether the example represents minutes, hours or days before proceeding. The mentor can ask for an updated definition and an explanation of how the choice affects the proposed code. Do not invent the correct unit to make the portfolio appear complete; the supplied fictional record does not identify it.

On small screens, scroll the table horizontally, or focus it and use the arrow keys.

5. Make feedback specific and credit actual contributions

Useful feedback identifies a location, explains the issue and defines what would address it. For the dictionary, a mentor might write: “The duration field has no unit. Confirm the intended unit in the exercise specification and revise the dictionary before running the script.” That does not pretend the mentor has checked an output or that the learner has made the correction. Retain the response and later review as separate dated events.

Agree contribution roles and update them when the work occurs. CRediT describes contributions and does not determine authorship. A planned analysis role is not an analysis already performed, and attending a mentoring session does not itself settle an author list. Keep contribution descriptions proportionate to the artifact and ask contributors to confirm them. Decisions about any eventual publication still need the applicable authorship criteria and the journal’s policies.

Reference: NISO CRediT: agreed contribution roles and authorship limits

6. Report progress without borrowing a future outcome

Use a short portfolio review to name the work inspected, the remaining dependency and the next agreed task. Compare revisions over time rather than counting files alone: a second file may be a corrected attempt, not a second completed competency. Ask the learner to explain an error in their own words and demonstrate the correction on the permitted exercise. Any claim of improvement should point to evidence that was actually reviewed.

For the fictional portfolio, the next proposed task is to resolve the dictionary unit under supervision. No real participant recruitment, institutional approval, analysis result, course completion or publication is documented. A team could later evaluate whether the mentoring sequence helps its learners, using a defined assessment plan. Until that evaluation exists, present the sequence as a practical proposal with inspectable tasks, not an effective programme already proven in practice.

Questions for a practical final check

• Each capability has an inspectable artifact.
• The practice dataset is explicitly synthetic.
• Course training and project permission have separate records.
• Draft code, execution and interpreted output remain distinct.
• Feedback points to a specific issue and review state.
• Contribution and progress statements describe work actually performed.