teacher profile teaching materials
The introductory unit briefly reconstructs the line leading from Skinner's analysis of teaching to programmed instruction, computer-assisted instruction and adaptive systems. This is not an erudite tribute. The categories of contingencies of reinforcement, immediate feedback, small-step progression and individual pace still constitute the implicit grammar of many intelligent systems, and recognising it allows students to assess more discerningly what is presented as radical innovation.
The core of the module then addresses three interwoven sets of questions. The first concerns how language models work and the documented techniques for governing their behaviour: here the main reference is Lee Boonstra's whitepaper for Google, adopted as an operational manual and critically discussed from a pedagogical standpoint. The second concerns the use of artificial intelligence in the tasks proper to education professionals: designing, adapting materials, building tests and rubrics, providing feedback, analysing learning data. The third concerns the use of artificial intelligence in educational research — from literature scoping to support for qualitative and quantitative analysis — with particular attention to the requirements of transparency, traceability and replicability demanded by the scientific community.
A concern for empirical verification runs through the entire course: every claim about the effectiveness of a tool is traced back to the available evidence, its methodological quality and its limitations.
1. **Genealogical premise: from teaching machines to adaptive systems**
- Experimental analysis of behaviour and teaching as the arrangement of contingencies of reinforcement.
- Immediate feedback, small steps, active participation, individual pace: four principles and their persistence in contemporary systems.
- From programmed instruction to computer-assisted instruction, e-learning environments, adaptive systems and intelligent tutoring systems.
- What actually changes with generative systems: from explicit rules to statistically learned behaviour.
2. **How language models work: the essentials required to use them well**
- Machine learning: task, experience, performance measure.
- Supervised and unsupervised learning, neural networks, transformer architectures.
- The model as a prediction engine: tokens, context, sequential generation.
- Output configuration: length, temperature, top-K, top-P; parameter interaction and consequences for reproducibility.
- Why a plausible output may be false: hallucinations, flattening of sources, repetition loops, opacity of internal processes.
3. **Prompt engineering: techniques, configurations and documentation**
- General and zero-shot prompting; one-shot and few-shot: number, quality and diversity of examples, handling of edge cases.
- System, contextual and role prompting: defining task, context and voice; overlaps and distinct functions.
- Step-back prompting: activating general knowledge before the specific task and mitigating bias.
- Chain of thought, self-consistency, tree of thoughts: intermediate reasoning, multiple paths, majority voting; computational costs and interpretive limits.
- ReAct and the use of external tools: a first approximation of agentic systems.
- Automatic prompt engineering: automated generation and scoring of variants; metrics and their inadequacy for educational tasks.
- Best practices: provide examples, prefer instructions over constraints, be specific about the output, design with simplicity, use variables, experiment with formats.
- Structured outputs and JSON schemas for both input and output; repairing malformed outputs.
- Systematic documentation of attempts: model, version, parameters, prompt, output, result, iteration. Documentation as a requirement of scientific rigour rather than a bureaucratic formality.
4. **Artificial intelligence and instructional design**
- From learning objective to prompt: translating educational intentions into operational requests.
- Generation and adaptation of materials: reformulation by readability level, exemplification, construction of cases and scenarios.
- Differentiation, accessibility and universal design for learning.
- Content sequencing, personalised pathways and their empirical basis.
- Specificities of adult learning: prior experience, motivation, self-direction, recognition of informal learning.
- What must not be delegated: pedagogical decision-making, the educational relationship, professional responsibility.
5. **Artificial intelligence and learning assessment**
- Assisted construction of multiple-choice items: plausible distractors, difficulty control, typical errors produced by models.
- Item analysis and human revision: difficulty and discrimination indices as quality control criteria.
- Analytic and holistic rubrics: generation, calibration, verification of descriptive coherence across levels.
- Automated feedback and formative assessment: potential, risks of standardisation, effects on motivation.
- Automated scoring of open-ended tasks: state of the evidence, agreement with human raters, fairness across subgroups.
- Validity, reliability and fairness: why automation does not exempt anyone from classical docimological requirements.
6. **Artificial intelligence in educational research**
- Literature scoping and organisation: querying, synthesis, risk of fabricated references, verification procedures.
- Retrieval-augmented generation (RAG) and the documentary grounding of responses.
- Support for qualitative research: assisted coding, comparison between human and automated coding, inter-coder agreement, risks of interpretive circularity.
- Support for quantitative research: data exploration, generation of analysis code, interpretation of results and its pitfalls.
- Educational Data Mining and learning analytics: predictive models, risk indicators, validity issues and self-fulfilling prophecy effects.
- AI-assisted scientific writing: language revision, synthesis, structuring; the boundary between legitimate support and the substitution of authorship.
- Transparency, traceability, replicability: disclosure statements, journal policies, positions of ethics committees.
7. **AI literacy, ethics, law and academic integrity**
- Competences required to understand, use, evaluate and critically discuss artificial intelligence systems; international frameworks.
- Algorithmic bias, stereotypes, inequalities of access, the human labour involved in training and moderation.
- Personal data protection, surveillance, learner autonomy, transparency and accountability.
- The European regulatory framework and the classification of systems in education.
- Hybrid writing, attribution, intellectual property; from plagiarism to postplagiarism.
- Limitations of automated detectors of generated text and the consequent redesign of assessment activities.
- Agrusti, F. (Ed.). (2023). *Educazione e Intelligenza Artificiale*. Rome: Roma TrE-Press. DOI: 10.13134/979-12-5977-257-2.
- Boonstra, L. (2025). *Prompt Engineering* [Whitepaper]. Google. Available at: https://www.kaggle.com/whitepaper-prompt-engineering. The document is adopted as the operational manual of the module and will be discussed critically from a pedagogical perspective; the required knowledge concerns prompting techniques, output configuration parameters, best practices and documentation procedures.
- Skinner, B. F. (1970). *La tecnologia dell'insegnamento*. Brescia: La Scuola. The following chapters and page ranges only, available on the course platform: Chapter 1, pp. 37-47; Chapter 2, pp. 48-69.
- Slides and learning materials provided by the lecturer through the course platform.
Mandatory readings must be studied in full for the specified sections. The applications presented during the lessons may change as tools evolve. The examination will focus on concepts, techniques, pedagogical criteria and procedures of analysis and verification rather than on the merely operational knowledge of a single commercial product.
Programme
The module aims to develop professional competence rather than mere familiarity of use. The distinction matters. Knowing how to query a generative system is not the same as knowing when to query it, with which technique, under which constraints and with which control procedures: it is this second order of competence that the course pursues.The introductory unit briefly reconstructs the line leading from Skinner's analysis of teaching to programmed instruction, computer-assisted instruction and adaptive systems. This is not an erudite tribute. The categories of contingencies of reinforcement, immediate feedback, small-step progression and individual pace still constitute the implicit grammar of many intelligent systems, and recognising it allows students to assess more discerningly what is presented as radical innovation.
The core of the module then addresses three interwoven sets of questions. The first concerns how language models work and the documented techniques for governing their behaviour: here the main reference is Lee Boonstra's whitepaper for Google, adopted as an operational manual and critically discussed from a pedagogical standpoint. The second concerns the use of artificial intelligence in the tasks proper to education professionals: designing, adapting materials, building tests and rubrics, providing feedback, analysing learning data. The third concerns the use of artificial intelligence in educational research — from literature scoping to support for qualitative and quantitative analysis — with particular attention to the requirements of transparency, traceability and replicability demanded by the scientific community.
A concern for empirical verification runs through the entire course: every claim about the effectiveness of a tool is traced back to the available evidence, its methodological quality and its limitations.
1. **Genealogical premise: from teaching machines to adaptive systems**
- Experimental analysis of behaviour and teaching as the arrangement of contingencies of reinforcement.
- Immediate feedback, small steps, active participation, individual pace: four principles and their persistence in contemporary systems.
- From programmed instruction to computer-assisted instruction, e-learning environments, adaptive systems and intelligent tutoring systems.
- What actually changes with generative systems: from explicit rules to statistically learned behaviour.
2. **How language models work: the essentials required to use them well**
- Machine learning: task, experience, performance measure.
- Supervised and unsupervised learning, neural networks, transformer architectures.
- The model as a prediction engine: tokens, context, sequential generation.
- Output configuration: length, temperature, top-K, top-P; parameter interaction and consequences for reproducibility.
- Why a plausible output may be false: hallucinations, flattening of sources, repetition loops, opacity of internal processes.
3. **Prompt engineering: techniques, configurations and documentation**
- General and zero-shot prompting; one-shot and few-shot: number, quality and diversity of examples, handling of edge cases.
- System, contextual and role prompting: defining task, context and voice; overlaps and distinct functions.
- Step-back prompting: activating general knowledge before the specific task and mitigating bias.
- Chain of thought, self-consistency, tree of thoughts: intermediate reasoning, multiple paths, majority voting; computational costs and interpretive limits.
- ReAct and the use of external tools: a first approximation of agentic systems.
- Automatic prompt engineering: automated generation and scoring of variants; metrics and their inadequacy for educational tasks.
- Best practices: provide examples, prefer instructions over constraints, be specific about the output, design with simplicity, use variables, experiment with formats.
- Structured outputs and JSON schemas for both input and output; repairing malformed outputs.
- Systematic documentation of attempts: model, version, parameters, prompt, output, result, iteration. Documentation as a requirement of scientific rigour rather than a bureaucratic formality.
4. **Artificial intelligence and instructional design**
- From learning objective to prompt: translating educational intentions into operational requests.
- Generation and adaptation of materials: reformulation by readability level, exemplification, construction of cases and scenarios.
- Differentiation, accessibility and universal design for learning.
- Content sequencing, personalised pathways and their empirical basis.
- Specificities of adult learning: prior experience, motivation, self-direction, recognition of informal learning.
- What must not be delegated: pedagogical decision-making, the educational relationship, professional responsibility.
5. **Artificial intelligence and learning assessment**
- Assisted construction of multiple-choice items: plausible distractors, difficulty control, typical errors produced by models.
- Item analysis and human revision: difficulty and discrimination indices as quality control criteria.
- Analytic and holistic rubrics: generation, calibration, verification of descriptive coherence across levels.
- Automated feedback and formative assessment: potential, risks of standardisation, effects on motivation.
- Automated scoring of open-ended tasks: state of the evidence, agreement with human raters, fairness across subgroups.
- Validity, reliability and fairness: why automation does not exempt anyone from classical docimological requirements.
6. **Artificial intelligence in educational research**
- Literature scoping and organisation: querying, synthesis, risk of fabricated references, verification procedures.
- Retrieval-augmented generation (RAG) and the documentary grounding of responses.
- Support for qualitative research: assisted coding, comparison between human and automated coding, inter-coder agreement, risks of interpretive circularity.
- Support for quantitative research: data exploration, generation of analysis code, interpretation of results and its pitfalls.
- Educational Data Mining and learning analytics: predictive models, risk indicators, validity issues and self-fulfilling prophecy effects.
- AI-assisted scientific writing: language revision, synthesis, structuring; the boundary between legitimate support and the substitution of authorship.
- Transparency, traceability, replicability: disclosure statements, journal policies, positions of ethics committees.
7. **AI literacy, ethics, law and academic integrity**
- Competences required to understand, use, evaluate and critically discuss artificial intelligence systems; international frameworks.
- Algorithmic bias, stereotypes, inequalities of access, the human labour involved in training and moderation.
- Personal data protection, surveillance, learner autonomy, transparency and accountability.
- The European regulatory framework and the classification of systems in education.
- Hybrid writing, attribution, intellectual property; from plagiarism to postplagiarism.
- Limitations of automated detectors of generated text and the consequent redesign of assessment activities.
Core Documentation
**Mandatory readings**- Agrusti, F. (Ed.). (2023). *Educazione e Intelligenza Artificiale*. Rome: Roma TrE-Press. DOI: 10.13134/979-12-5977-257-2.
- Boonstra, L. (2025). *Prompt Engineering* [Whitepaper]. Google. Available at: https://www.kaggle.com/whitepaper-prompt-engineering. The document is adopted as the operational manual of the module and will be discussed critically from a pedagogical perspective; the required knowledge concerns prompting techniques, output configuration parameters, best practices and documentation procedures.
- Skinner, B. F. (1970). *La tecnologia dell'insegnamento*. Brescia: La Scuola. The following chapters and page ranges only, available on the course platform: Chapter 1, pp. 37-47; Chapter 2, pp. 48-69.
- Slides and learning materials provided by the lecturer through the course platform.
Mandatory readings must be studied in full for the specified sections. The applications presented during the lessons may change as tools evolve. The examination will focus on concepts, techniques, pedagogical criteria and procedures of analysis and verification rather than on the merely operational knowledge of a single commercial product.
Reference Bibliography
**Further reading** - Eaton, S. E. (2023). Postplagiarism: transdisciplinary ethics and integrity in the age of artificial intelligence and neurotechnology. *International Journal for Educational Integrity, 19*, 23. https://doi.org/10.1007/s40979-023-00144-1 - UNESCO. (2024). *AI competency framework for teachers*. Paris: UNESCO. https://doi.org/10.54675/ZJTE2084 - European Commission, Directorate-General for Education, Youth, Sport and Culture. (2022). *Ethical guidelines on the use of artificial intelligence and data in teaching and learning for educators*. Publications Office of the European Union. https://data.europa.eu/doi/10.2766/153756 - Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence. Limited to the provisions relevant to systems used in education and training. - Scientific articles, institutional reports and case studies on artificial intelligence in education indicated during the course. Further readings are not mandatory unless otherwise stated by the lecturer. They are recommended for both attending and non-attending students.Attendance
FacultativeType of evaluation
**Formative assessment.** Lessons will include structured self-assessment tests, exercises in the comparative analysis of generated outputs, peer review of documented prompts, and short design activities. These activities will not directly contribute to the final grade. They will allow students to monitor their preparation and receive feedback on errors and adopted strategies. **Final assessment.** The assessment of the module forms part of the overall examination for *Experimental Pedagogy: Models and Procedures for Adult Education*. The module will be assessed through multiple-choice questions designed to evaluate conceptual knowledge, understanding of the relationship between prompting techniques and educational tasks, command of output verification criteria, and the ability to critically assess uses of artificial intelligence in teaching and research. Incorrect answers will not incur a penalty. Grades will be determined through both norm-referenced and criterion-referenced procedures, consistently with the assessment arrangements for the full course.