On this page

A free edited volume from the Society for the Teaching of Psychology (2026, ISBN 978-1-941804-82-7) that organizes generative AI course by course rather than tool by tool. Psychology supplies the respondents to many AI-use surveys without being their subject; this book is the counter-case, with 21 course chapters plus an introduction and a conclusion across 255 pages.

The argument the book makes

The editors' premise is that psychology instructors are unusually well placed to teach AI use, because the discipline already teaches judgment about measurement, evidence and human behavior. Their claim is narrower than "AI literacy matters". It is that AI literacy is developmental: students do not have it or lack it, they acquire it through staged access, reflection and feedback.

Three design commitments follow, and most chapters enact at least one of them. Stage what students may do with a tool, moving from guided use toward independent use as judgment develops. Pair use with verification, so a claim made with a model has to be checked against a source and shown. Define success as judgment about when and whether to use the tool, not fluency with it.

The structure is the argument. Chapters are grouped by the course an instructor is about to teach — foundational and cross-cutting, clinical and counseling, research methods and statistics — rather than by tool, capability or institutional policy.

Section I: foundational and cross-cutting courses

  • Access is the first problem, not the tool (Slade and Hyk). Students who arrive without prior AI experience are disadvantaged twice: they lack the prompting skill and they lack the judgment about when a fluent answer is wrong. The chapter teaches prompt engineering alongside open repositories of reusable prompts, and treats that instruction as an equity measure. The author discloses a financial interest in a company commercializing educational AI.
  • AI as a content generator for teaching (Corcoran). The chapter's use case is the instructor's own preparation: visuals, examples, problems and other teaching stimuli generated against a stated objective. Its argument is that customization to that objective, not novelty, is what makes the output useful. The chapter also insists that students be taught about AI — its limitations and ethics — and not merely handed it.
  • Study skills as the bridge to theory (Costa). In an upper-level learning and memory course, students use AI study tools and then map what the tools do onto spaced practice and retrieval. The design makes the psychology the subject and the tool the specimen, which is the cleanest way to teach a course whose content is the learning science the tools claim to implement.
  • An applied behavior-change project (Emanuel). Students run a health behavior change project with an AI system as coach. Their reflections separate what worked — immediate, tailored, non-judgmental suggestions — from what did not: no empathy, and plausible misinformation about health. The chapter is one of the few that puts a model in a role where being wrong carries a cost.
  • AI as debate coach (Jackson). An upper-level social psychology course uses a model to map arguments, surface counterarguments and rehearse rebuttals before a public debate, with every claim verified against peer-reviewed sources. Of the written reflections, 75% found the model at least moderately useful in preparation and 35% were strongly positive; about 10% found little use, mostly because the model produced generic arguments or confident errors. The later iteration adds a one-page prep plan and an evidence ledger linking each claim to a source.
  • A course chatbot with feedback duties (Lee, Nam and Wages). A chatbot built for Introduction to Psychology, Research Methods and Psychology Apprenticeship supported topic brainstorming, research question refinement and paper structure. The authors report 124 students in the pre-test survey and a 25% post-test response rate, so the outcome evidence is thinner than the implementation detail.
  • Assessment in three tiers (Granito). Diagnostic, formative and summative tasks each get an AI-enhanced version, and the summative end moves toward authentic projects where the tool is a support partner rather than the deliverable. This is the chapter to read if the question is what to stop grading.

Section II: clinical and counseling courses

  • Psychotherapy as an introductory activity (Castro). Students pick a DSM-5-TR disorder and a psychotherapy technique, then experience a simulated session and discuss what AI-based telehealth gains and loses. The design teaches the technique by contrast: the simulation shows what the relational part of therapy is by omitting it.
  • Building case studies together (Pulido). Students use ChatGPT to draft DSM-5-TR case studies and then evaluate them, which puts the assessment skill on the student rather than the model. The chapter describes graduate and undergraduate adaptations and treats iterative prompt refinement as part of the clinical reasoning, not a technical chore.
  • Case conceptualization in a seminar (Draheim). Case conceptualization is where students synthesize client information and treatment outcome research, and it is hard to teach at undergraduate level. The seminar uses AI to generate and critique conceptualizations, with the instructor's feedback load named as the constraint that makes the tool attractive in the first place.
  • Cultural humility (Draheim). A homework assignment uses a model to work through an article on cultural humility, followed by an in-class exercise in finding the biases inside generative output. The pairing matters: the tool is used to open a topic and then made the object of critique in the same week.
  • Trauma-responsive teaching (Rubio). Graduate trauma courses use text and image models to co-create therapeutic stories and safe-place imagery, grounded in Herman's tri-phasic model and Perry's neurosequential model. The framework does the work here; the tools supply material the framework then constrains.
  • Bots as therapists (Niederjohn). Students talk to several chatbots and compare the experience with face-to-face work on therapeutic alliance. The chapter supplies the numbers that make the exercise urgent: 65% of mental health professionals in earlier surveys saw benefits in chatbot support, while 86% did not support their use in care. Its practical suggestion is a ladder — AI role-play first, then peers, then human simulated patients.
  • Role-play for trainees (Labriola and VanHara). AI role-plays replace scripted peer role-play with unscripted clinical presentations, which the authors argue raises realism and willingness to take risks. They anchor the need in student survey data: 86% of nearly 4,000 students across 16 countries reported using AI, 58% said they lacked sufficient knowledge of it, and 72% wanted courses in AI literacy.

Section III: research methods, statistics and scientific literacy

  • AI as research co-pilot (Sumutka and colleagues). Activities walk novices through narrowing a topic, finding sources, generating and operationalizing variables, and applying ethics. The stated goal is information literacy first and speed second, and the chapter is explicit that the tool is being modeled as much as used.
  • Question generation as critical evaluation (Burkett). A three-phase approach in an undergraduate research methods course turns ChatGPT into a generator of evaluative questions about study designs, with a custom GPT built for the purpose. The questions then serve as a measure of understanding — an assessment move that other chapters talk about but rarely operationalize.
  • Uses and abuses, as modules (Urdan, Firtch and White). Ready-to-use modules cover research design generation, practice tests and the known failure modes, on the reasoning that adoption friction is the real barrier for instructors who are AI-curious but time-poor.
  • The chatbot that fell short (Hilley and Edwards). "Zig", built on the Juji platform for a required statistics course, was meant to give on-demand help with practice problems. It did not meet expectations. Its failures are specific and instructive: a student asking about Levene's test typed "levee" and got a refusal, a question about grading produced an unrelated answer about z-scores, and requests for examples got definitions instead. Much of this traces to a Q&A-loaded platform rather than a reasoning model, but the chapter is the volume's honest negative result and the one to read before building a course bot.
  • Staged integration in R teaching (Garrison, Tyson and Lyu). The four-stage model begins with AI-free learning and adds tools only as students can evaluate what the tools produce. The readiness signal the authors name is frustration with AI's limits, and the failure mode they document is introducing AI earlier, which produced dependency before the fundamentals were there.
  • ChatGPT plus Elicit in a senior seminar (Katzarska-Miller). Four preparatory workshops precede a semester-long research paper, with a generative model for drafting and a research assistant tool for literature. The chapter grounds the design in social constructionist learning theory and uses pre- and post-course self-assessment to check whether AI literacy moved.

What recurs across the twenty-one chapters

  • Stage the access. Nearly every chapter withholds the tool until students have something to judge it against, whether that is a statistic, a framework, a debate position or a programming fundamental.
  • Require verification and make it visible. Evidence ledgers, source checks, prompt logs and "what changed after fact-checking" notes appear across sections. The requirement is what converts tool use into learning.
  • Use student reflection as the instrument. Reflections are the evidence base for most chapters, which is also the volume's central methodological weakness.
  • Design at course level, not tool level. The book's own organizing decision — one chapter per course — is the recommendation.
  • Keep the instructor's judgment as the object. The recurring claim is that students should leave able to decide when not to use the tool.

Where the book is thin

The chapters are practitioner accounts. Almost none report a design, a sample or an effect size, and where numbers appear they describe implementation rather than learning. The exceptions are partial: Lee and colleagues have pre- and post-test surveys with a 25% post-test response rate, and Jackson's reflection analysis is a thematic coding of one class.

The enthusiasm is also unevenly checked. Corcoran and Costa write as advocates; Hilley and Edwards report a failure; only Slade and Hyk disclose a commercial interest. Chapters naming specific products will date quickly, since most describe tools as they behaved during 2025.

Coverage gaps worth knowing: cost and premium tiers are barely discussed, accessibility and disability are not treated as a section, and K-12 appears only in Castro's secondary-education chapter. The volume is undergraduate and graduate psychology teaching in the United States, with little on large lecture courses beyond Costa's and Granito's.

What you can do with it

Read the chapter for the course you are about to teach. The three sections — foundational and cross-cutting courses, clinical and counseling courses, research methods and statistics — keep related chapters together. Examples: an AI tutor chatbot in psychological statistics, ChatGPT paired with Elicit in a senior seminar, AI as a debate coach, case conceptualization and cultural humility in clinical training, and a research methods chapter that has students generate questions with ChatGPT and then critique them.

Who it is for

Psychology instructors at any career stage, and instructors in adjacent social science disciplines whose assessment problems have the same shape: essay-heavy work, case discussion, statistics anxiety, small courses where one redesign is feasible.

Notes and caveats

Free PDF, no account. The copyright is STP's, not a Creative Commons license: personal and classroom copies are permitted with attribution, while redistribution or commercial use needs the author's permission. Chapters are practitioner accounts, not studies — no designs, samples or effect sizes — so treat claims as experience reports and cite chapters individually.

Connected Concepts

Embed this page

Copy the code below to embed this page in a learning management system or other website. The embedded version shows only the page content — it hides the site header, navigation, and footer.