Data and AI Literacy in Geography Education: A Needs Assessment of Pre-Service Geography Teachers

Synthesis: Rink and Peter ask what pre-service geography teachers in Germany believe they need in order to teach with data and with AI. Eighteen advanced trainees — nine university teacher-training students and nine trainee teachers, mean age 26.33 — gave semi-structured interviews of at least 60 minutes, coded in MAXQDA through a deductive-inductive qualitative content analysis that produced 76 categories. Needs clustered overwhelmingly under teacher education rather than school conditions: 18 of 18 named data-related education needs and 17 of 18 named AI-related ones. Data needs were broad and generic — literacy, handling, visualization, evaluation — while AI needs were more subject-specific, centered on content-related AI knowledge. Ethical and critical dimensions were rarely raised, a gap the authors treat as significant.

Key Findings

1. Teacher education absorbed almost all the need. All 18 participants described data-related needs in teacher education against 4 who mentioned school conditions; for AI the split was 17 against 1.
2. Data handling outweighed digitalisation. Within teacher education, 18 raised data handling and 9 raised digitalisation; data literacy was named by 16, technological pedagogical data knowledge by 12 and pedagogical content data knowledge by 11.
3. The data needs were general rather than geographical. Only 1 participant raised technological content data knowledge, and geography-specific tools such as GIS appeared rarely, though 5 named acquisition and 4 each named visualization and evaluation.
4. AI needs were more subject-specific. Content-related AI knowledge was the largest AI category (n = 11), with 6 wanting to know how AI applies to geography topics, ahead of pedagogical AI knowledge (n = 9) and knowledge of AI tools and their implications (n = 5).
5. Handling student AI use was a minority concern. Pedagogical content AI knowledge drew the fewest mentions, including awareness of students' AI use (n = 3) and data manipulation with AI (n = 1); 2 participants wanted help with AI-generated task solutions.
6. Ethics barely surfaced. The authors report that safe, ethical and critical use of AI was "only marginally referenced, if at all", and that offering students a critical perspective on AI was rarely mentioned.

Design, sample and coding

This is a qualitative needs assessment set in Hesse, Germany. Participants were recruited through a request to all teacher-training institutes in the federal state and through geography teacher-training students at Marburg University; data were collected in 2025 in 16 online and 2 in-person interviews of at least 60 minutes each. The sample was deliberately balanced: 9 women and 9 men, mean age 26.33, split evenly between university teacher-training students and trainee teachers in the practical phase. The inclusion criterion was advanced training — university students had to have completed a supervised school internship — so that everyone had planned and taught geography lessons and could speak to their own needs.

The guide followed the principle "as open as possible, as structured as necessary" and was refined through three pretests before being split into a data part and an AI part matching the research questions. Analysis used a seven-step structuring qualitative content analysis in MAXQDA 24. Two main categories were derived deductively; subcategories were built both deductively, for instance from Grillenberger and Romeike's data life-cycle, and inductively, producing 76 categories. Two coders worked through the full material with an agreed frame, a third reviewed 15 transcripts, and disagreements were resolved in discussion. Each category was counted once per participant, so frequencies top out at 18.

Two framework choices shape the results. Data categories borrow from the TPACK model, renamed to speak of data rather than technology, and from DPACK for digitalisation. AI categories borrow from AI-PACK, plus an inductively added category for knowledge of AI tools and their implications. The terminology was adjusted because the study asked about needs, not competence gaps.

Data: strong demand, weak subject specificity

Under teacher education, data handling drew all 18 participants and digitalisation 9. Within data handling the largest single need was general handling (n = 7), followed by acquisition (n = 5), visualization (n = 4), evaluation (n = 4) and an explicit call for data literacy (n = 4). The quotations are practical: where to find usable maps and datasets, how to judge whether a source is appropriate, how to read a data model. One participant asked how to "assess the data I can use"; another said "as a teacher, it can be difficult to determine where to find relevant data".

The geography-specific demand sits in the pedagogical layer, and it is thinner than the general one. Twelve participants raised technological pedagogical data knowledge, and within it 7 wanted to learn how to teach data literacy — helping students critically analyze and evaluate data — while 6 wanted guidance on which data suit which grade levels and levels of complexity. Eleven raised pedagogical content data knowledge, chiefly 8 who wanted examples of how to incorporate data into geography lessons and 5 who wanted general guidance on handling data in lessons. Eight raised content data knowledge, mostly around why data matter and how to connect them to real-world issues.

What is missing is as telling as what is present. Technological content data knowledge drew a single mention, and geospatial technologies, GIS and place-based tools — the discipline's distinctive contribution in the literature — appear mainly as illustrations of a general digital-tools need. The authors read this against Muschaweck's finding that pre-service geography teachers need topic-specific TPACK and Urfan et al.'s result that technological knowledge lags other domains in this group, concluding that the need sits in the pedagogical context rather than the technical content context.

AI: content knowledge first, critique last

The AI results invert the data pattern. Where data needs were general, AI needs were comparatively subject-specific. Content-related AI knowledge was the largest category (n = 11): 6 participants wanted to understand how AI can be used in a geography-specific context, 3 wanted help designing appropriate tasks, and 3 raised risks attached to content not generally associated with AI tools. Pedagogical AI knowledge followed with 9 participants wanting to see what AI use might look like in a classroom, ideally through demonstration modules. Knowledge of AI tools and their implications drew 5, of whom 4 wanted concrete examples.

One quotation captures the uncertainty: "There's definitely a need for further training because this is all so new and uncertain. We don't even know what it can or can't do or how it can be used." That is a need for orientation as much as for technique, and it sits awkwardly beside the finding that pedagogical content AI knowledge was the smallest category.

The phase findings are small but consistent. Seven participants tied their needs to a stage of teacher education: 3 said AI should be part of university study, while continuing professional development drew 2 calls for mandatory training and 2 for structured exchange with subject colleagues. One participant raised school conditions, noting that many schools lack devices for students as well as teachers.

The authors are explicit about the ethical gap. Measured against AI literacy frameworks that include ethics and critical evaluation, their participants' needs concentrate on knowledge and use. "Understand", in Mills et al.'s sense, referred to how AI can be applied in geography teaching rather than to how it works, and critical evaluation appeared mainly as reflection on impact and data manipulation. Safe and ethical use in schools was, in their words, only marginally referenced. They offer four explanations without choosing among them: lower perceived relevance, career stage, limited awareness of the gap, or the limits of self-report.

What this means for practice

• Teacher educators. Build data content into geography didactics around the data life-cycle — acquisition, visualization, evaluation, archiving — rather than assuming general digital-tools coverage transfers.
• Program designers. Pair AI content with subject examples: participants wanted to see AI applied to geographical phenomena and to practice judging appropriate tasks, which is content knowledge work rather than a generic literacy module.
• Faculty and school leaders. Make ethics and critical evaluation an explicit teaching objective, because participants did not raise them unprompted and a needs assessment cannot surface a gap the respondent does not perceive.

Limitations

• The sample is 18 pre-service teachers from one German federal state, so findings generalize analytically rather than statistically.
• Needs are self-reported: participants can only name gaps they are aware of, and the design cannot infer from silence.
• The study reports needs, not measured competence; frequencies indicate salience, not severity.
• Only representative quotations appear, and the German-to-English translation adds an interpretive step.
• University students and trainee teachers were pooled rather than compared.

Citation

Rink, L. S., & Peter, C. (2026). Data and AI Literacy in Geography Education: A Needs Assessment of Pre-Service Geography Teachers. Education Sciences, 16(11), 1824.