Assessing transversal skills (communication, collaboration, critical thinking, time management) is a stated priority in high schools and universities, but it remains one of the most delicate tasks for those who teach. On the one hand, students and institutions ask for evidence: “What is this judgment based on?”. On the other, teachers must protectacademic security(assessment integrity, decision traceability, consistency of criteria) in a context in whichartificial intelligenceis changing both study practices and expectations around assessment. In this article we propose a didactic-operational approach: how to maketransversal skills assessmentmore defensible, fairer, and more sustainable, and how a solution likeStudierAIcan support a more robust assessment framework, with specific attention to academic security. If you want to explore the approach in a practical way, you can alsostart for freeand test an assessment workflow on one of your activities.
Why assessing transversal skills has become central (and difficult)
In the transition from school to work (and, at university, from the course to the profession), transversal skills are often the factor that determines performance quality more than disciplinary knowledge alone. Being able to justify a choice, collaborate effectively, manage uncertainty, make responsible decisions: these are abilities that emerge in authentic contexts, not in a single multiple-choice question.
From a pedagogical standpoint, this centrality is not a fad. The literature onauthentic assessment, situated learning, and analytic rubrics shows that complex competencies require: (1) meaningful tasks, (2) explicit criteria, (3) observable evidence, and (4) improvement-oriented feedback. The problem is that these four elements, in day-to-day practice, are difficult to keep together coherently.
Why is it difficult? Because soft skills are: multidimensional (multiple indicators for the same competence), context-dependent (good communication in a lab is not the same as in a seminar), and often distributed over time (they show up across multiple moments, not in a single assessment). In addition, in classrooms and universities the need is growing to make assessmentdefensible: being able to explain criteria, steps, and sources of evidence, especially when assessment affects credits, certifications, or access to selective pathways.
Critical issues with traditional methods: subjectivity, teacher workload, and poor traceability
The most common practices for assessing transversal skills include rubrics, ongoing observations, learning logs, peer assessment, and authentic assessments (project work, presentations, cases). They are valid tools, but they have recurring limitations that are amplified when groups are large or when multiple teachers assess the same learning outcome.
Three critical issues deserve attention because they directly impact equity and quality:
- Subjectivity and bias: even with well-written rubrics, interpretation of descriptors can vary. Effects such as the “halo effect”, severity/leniency, and prior expectations influence judgment, especially when evidence is not collected systematically.
- Teacher workload: observing, taking notes, synthesizing, and providing feedback takes time. In many cases, assessment of transversal skills ends up being “compressed” into a few moments, reducing data validity and feedback quality.
- Poor traceability: often only a final grade and a few notes remain. If a student asks for clarification (or if the institution must demonstrate consistency and correctness), it is difficult to reconstruct the decision-making path: which evidence, which criteria, which weights.
These limits do not “invalidate” traditional methods, but they indicate a need: to better standardize criteria, collect evidence more systematically, and reduce unwanted variability. In terms of assessment quality, the goal is not to eliminate the teacher’s professional judgment, but to make it more consistent and supported by observable data.
How artificial intelligence can make assessment more reliable and automated
When talking about AI in assessment, it is useful to distinguish between two uses: (1) AI as a shortcut to “generate answers” (risky for integrity and learning) and (2) AI as infrastructure to make assessment more rigorous, traceable, and repeatable. In the second case, AI does not replace the teacher: it automates parts of the process that are currently manual and fragile (evidence collection, consistent application of criteria, report production), leaving the teacher in charge of the instructional design and the final decision when needed.
A credible AI approach to assessing transversal skills is based on a few methodological principles:
- Standardized scenarios and tasks: comparable problem situations, with clear constraints and expected outputs (e.g., a short argument, an action plan, negotiation of roles in a team).
- Explicit criteria and anchored rubrics: observable descriptors (not impressions) and levels with performance examples. This reduces ambiguity and improves alignment among teachers.
- Evidence analysis: AI can identify indicators in texts, activity logs, discussion contributions, or project outputs (always with respect for privacy), linking them to the rubric criteria.
- Standardized reporting and auditability: comparable summaries across students and across course editions, with a record of which evidence supported each level assignment.
On the instructional side, the most immediate advantage is the ability to increase feedback frequency without proportionally increasing teacher workload. If evidence collection and pre-analysis are automated, the teacher can focus on what makes the difference: discussing criteria with the class, designing authentic tasks, intervening in a targeted way on students’ needs.
To be reliable, however, AI must be embedded in a governance framework: transparency of criteria, teacher oversight, data management, and measures to prevent abuse. This is where academic security becomes a design requirement, not an afterthought.
StudierAI: secure testing, certification, and reporting of transversal skills
In a context where soft skills assessment risks being uneven,StudierAIpositions itself as support forinnovative teachingwith an explicit focus on standardization, evidence, and academic security. The key idea is to transform assessment from a “final impression” into a documented process: structured tasks, shared rubrics, evidence collection, and consistent reporting across classes or courses.
For a teacher, this translates into a few operational possibilities:
- Structured competency-based assessments: scenarios that require justified decisions, constraint management, effective communication, and collaboration (including asynchronously).
- Shared rubrics and comparability: explicit criteria that help maintain consistency across teachers, parallel classes, or different modules of the same course.
- Evidence generation and organization: orderly collection of student outputs (texts, scenario responses, contributions) and linking them to assessment criteria.
- Reporting and certification: readable summaries for students and institutions, useful for guidance, tutoring, recognition of competencies, and improvement pathways.
The word “security” here should be understood broadly: not only data protection, but alsoacademic integrityand the robustness of the assessment process. A system that is useful to teachers should help to: reduce the possibility of assessment manipulation, maintain a record of decisions (audit trail), and make criteria clear before the assessment. This increases transparency as perceived by students and reduces instructional disputes.
An often underestimated aspect is the formative effect of assessment: when students see clear criteria and consistent feedback, they tend to invest more in the quality of the process (for example, revising an assignment or negotiating roles in a group) rather than only optimizing the grade. In other words: more traceable assessment also improves learning, not just the administration of judgment.
If you want to evaluate the tool’s suitability for your context, the best way is to start from a real task in your course and check: (1) how clear the criteria are, (2) how observable the evidence is, (3) how sustainable the workflow is. You cansign up for freeand set up a small pilot, or learn more about the project’s approach and values on theabout uspage.
Implementation in the classroom and at university: best practices, governance, and success criteria
Introducing AI tools for assessment works when it is treated as an instructional and organizational project, not as a simple “tool”. Below is an essential roadmap, adaptable to high school and university, to integrate a transversal skills assessment system while keeping equity and academic security at the center.
1) Design a targeted pilot (2–6 weeks). Choose just one priority transversal skill (e.g., argumentative communication or collaboration) and an authentic task already present in your course. Define a short rubric (3–5 criteria) with observable descriptors. The goal of the pilot is not to “measure everything”, but to verify feasibility and evidence quality.
2) Train teachers and tutors on the criteria, not only on the tool. Effective training includes examples of work at different levels, calibration exercises among raters, and discussion of ambiguous cases. This step reduces inter-teacher variability and makes AI a support for consistency, not an opaque referee.
3) Define usage policies and communicate them to students. Explain transparently: what data are used, for what purpose, how criteria are applied, and what room remains for the teacher’s judgment. Include guidance on permitted student use of generative tools (if applicable), distinguishing between study support and assessed production. Preventive clarity is a powerful lever for academic security.
4) Integrate with the LMS and existing workflows. Where possible, connect evidence collection to assignments already in place (homework, forums, project work). The goal is to avoid “double entry” and reduce operational friction. Even without technical integrations, it is useful to standardize formats and assignment naming to improve traceability.
5) Plan a review and improvement cycle. After the pilot, analyze where the rubric proved too generic, which criteria generated the most disagreement, and which evidence was most informative. Update the task and rubric before scaling to more classes or courses.
To understand whether implementation is working, you need concrete indicators (KPIs) that combine instructional effectiveness, equity, and security. Some practical examples:
- Reliability/consistency: reduced variability among raters on common samples; stability of results across course editions with the same criteria.
- Sustainability: average grading/feedback time per student; number of formative feedback instances delivered during the course (not only at the end of the module).
- Equity: analysis of any systematic gaps between groups (if relevant and in line with policies); monitoring of disputes and requests for clarification.
- Acceptance: students’ perception of transparency; perceived usefulness of feedback; clarity of criteria before the assessment.
- Academic security: presence of an audit trail for decisions; percentage of assessments with complete evidence; consistent handling of cases of non-compliance with usage rules.
In summary, AI can become an ally of assessment only if it is used to make criteria explicit, evidence systematic, and decisions traceable. This is the meeting point betweentransversal skills assessment, innovative teaching, and academic security: not “automating the grade”, but building a fairer, more formative, and more defensible process. With tools like StudierAI, the teacher can maintain pedagogical leadership and, at the same time, reduce operational workload and inconsistency, increasing feedback quality and trust in the assessment system.
