From Mastery to Judgement: Rethinking AI Professional Development in Language Education

If you are close to my age, you may remember a time when photographs took a while to develop. There was the anticipation while the film was being processed; the mild unease when you finally picked up the envelope, wondering what had come out well and what had not; and, often, the quiet satisfaction of recognising a holiday or a gathering¹ that you had enjoyed. Something of that emotional sequence came back to me as I read the participants’ reports from the workshop that the Artificial Intelligence in Language Education (AI Lang) project team organised at the end of November.

In this post, I would like to share my initial reaction to this rich input. It’s not so much an evaluation done because I am required to do one (even though we are preparing one of those as well). Rather, it’s my own personal attempt to see the workshop through the participants’ eyes, and to understand how they made sense of the activities we did together. My reason for sharing this is not to deliver a verdict on the workshop, but rather to explore the value in thinking aloud about what evaluation data can reveal (and, perhaps, what it leaves unresolved) when examining professional learning about AI and beyond. With this in mind, let me take you along as I try to make sense of what participants shared.

The AI Lang workshop

As a brief reminder, the AI Lang project is a four-year initiative (now in its third year) that aims to empower teachers to use Artificial Intelligence in language education responsibly and ethically. In the project, which is supported by the European Centre for Modern Languages (ECML) of the Council of Europe, we are developing a set of guidelines for using artificial intelligence, as well as a Moodle-based professional development course that enacts these guidelines. At roughly the midpoint of the project, on 26–27^th November 2025, we invited approximately 40 experts from all Council of Europe member states to Graz, where we shared our developing outputs with them and invited critical feedback. The evaluation discussed here draws on participants’ reflections from that event.

The Artificial Intelligence in Language Education (AI Lang) workshop

After the event, the ECML invited participants to produce brief reports outlining their experiences and main take-home messages. These written reflections tend to be particularly useful for documenting people’s perceptions and uncovering patterns in how they made sense of the event; they also work well with other forms of evidence (e.g., metrics) that measure impact in the more causal sense of the word. So, what I did in the last few days was a careful reading of the reports alongside one another as I tried to trace the patterns that emerged.

Learning as pedagogical confidence, not technical mastery

When reading evaluation data, what’s missing is often just as telling as what is in there. For me, one thing that stood out immediately was that very few participants described the workshop as being primarily about AI tools: there was remarkably little written about “learning the latest tools” or “finding out how to teach x with AI”. Nobody claimed they had “mastered AI” or become an expert user.

Instead, learning was framed in quieter but more durable terms, like “increased confidence”, “clearer” judgment, heightened “awareness of ethical and social dimensions of AI” and a better sense of the “limits as well as possibilities” of AI-assisted language education.

This matters, I think, because it represents a shift from technical competence (the technology of AI-supported language education) to teaching and learning implications (its methodology).² As one participant put it, the workshop helped them focus on “using AI safely, meaningfully, and with clearer pedagogical purpose,” rather than adopting tools for the sake of innovation. In other words, participants valued learning not just how to use AI, but whether and when it should be used at all.

Several participants explicitly contrasted this with the tool-focused training they’d experienced elsewhere. What they appreciated was the absence of pressure to “keep up” or “innovate fast.” The reflective pace felt empowering rather than overwhelming.

This raises a question: if the dominant paradigm in AI professional development prioritises technical skill acquisition, what gets lost?³ The participants’ responses suggest that space for critical judgment, i.e., the capacity to evaluate, adopt, adapt or resist, may be precisely what teachers need most.

If the dominant paradigm in AI professional development prioritizes technical skill acquisition, what gets lost?

Guidelines as a shared language

The draft AI Lang Guidelines emerged as a central anchor across the reports, regardless of participants’ national or institutional contexts. What reassured me was how they were described: not as checklists or compliance documents, but as “a framework that genuinely supports meaningful AI use in language education.”

A number of reports emphasised how the guidelines helped them structure conversations with colleagues:

Working in small groups of 7–8 participants, we carefully analysed the wording of each principle, debating their conceptual foundations, refining definitions, and proposing concrete behaviours or descriptors that could demonstrate adherence to these principles in everyday teaching practice.

In contexts where AI discussions can be politically or institutionally sensitive, this matters. The guidelines offer a shared language for framing conversations, structuring teacher training, and justifying decisions to school leadership or authorities. They create space for deliberation rather than dictating outcomes.

Moodle as a space for thinking, not just training

The second major output of the AI Lang project, the Moodle course, was also received in an interesting way.

In many of the reports, the participants repeatedly acknowledged that this is not a finished training package, but still referred to it as a resource that supports exploration, reflection, and collaboration, rather than as a finished training package. Some emphasised that the Moodle space made it possible to work with concrete examples while linking them to broader principles. In the words of one participant:

Beyond simply presenting AI tools, it offered concrete pedagogical ideas and examples that demonstrate how these tools can be integrated meaningfully and effectively into language teaching and learning.

The opportunity to contribute actively to the development of the resource was something many participants seemed to value (“It was equally valuable to contribute to […] the Moodle site”, wrote one participant), and they also commented positively on the “excellent opportunity” to reuse and adapt it to different professional contexts.

This suggests something about the relationship between polish and utility in professional learning resources. Perhaps unexpectedly, the course’s “unfinished” quality (a source of much unease for me before the workshop!) seemed to encourage participants to engage more.

From participants to multipliers

Perhaps the strongest signal in the reports was the evidence of professional impact. Several reports moved quickly from individual learning to supporting others, and they seemed to use language that emphasises guidance rather than advocacy. So, for example, someone wrote that “Sharing a European perspective gave me confidence to advise colleagues”, and another said that they would “share the materials and ideas […] with colleagues and help them approach the use of AI in a more informed and responsible way”.

In fact, most participants described very concrete plans for dissemination: mentoring colleagues, embedding ideas in teacher education programmes, influencing institutional digital plans, contributing to professional associations, or writing practitioner-oriented articles. Some even described themselves —unprompted— as acting as ambassadors for the work, a gesture that we all appreciate greatly.⁴

This matters because it suggests the workshop’s effects are distributed, delayed, and relational. They’re travelling through professional networks in ways we can’t directly observe or measure. The impact is multiplicative rather than additive, which makes it both more significant and harder to capture in conventional evaluation frameworks.

A workshop that resisted easy answers

If there’s one thread running through all the reports, it’s this: participants didn’t leave Graz believing that AI in language education is “solved.” They left with (I think) better questions about ethics, justice, sustainability, teacher workload, learner agency, and the role of human judgment in increasingly automated environments.

I, too, am left with better questions: What happens when teachers return to institutional contexts that reward rapid AI adoption over careful deliberation? How do we support the ongoing work of ethical judgment when the technological landscape keeps shifting? Who should make these decisions, if teachers feel overwhelmed by the responsibility? How do we measure professional learning that emphasises hesitation, critique, and restraint alongside capability?

The workshop captured a moment in an evolving conversation, and the rich reports that the participants submitted record where we are now, not where we (this is a very inclusive ‘we’) might end up.

A final thought

Evaluation itself isn’t a concluding act. It’s another form of professional learning. It is an occasion to pause, examine what happened, and consider what it might mean. If readers involved in AI-related professional development recognise some of these tensions in their own contexts, that recognition may be as valuable as any formal conclusion.

A luta continua.

---

---