Data Practicum: a Sequence of Courses on AI-First Applied Data Analysis
Barton Poulson, PhD, Professor of Psychological Science
Psychology and Counseling Department, Utah Valley University
Contents subject to frequent revision. Updated 2026-11-09
The Goals of the Data Practicum Courses
The nature of data work has changed dramatically since generative AI became generally available. In particular, generative AI is increasingly able to take on the technical skills of data work – how to write statistical code, how to choose and implement functions – while human analysts are increasingly able to devote more of their attention to the context and application of data work. These contextual skills – also known as “soft skills” or, better, “durable skills” because they do not become obsolete quickly – are critically important to the success of data projects. The happy result of the AI revolution is that, when it comes to data work, humans are able to focus on what humans do best, and are able to collaborate with AI on what it does best.
In an attempt to help students and professionals learn to work better with AI and hone their own, irreplaceable skills, I have proposed a series of 3 short courses that I am calling the “data practicum on AI-first data work.” In professional education, a “practicum” is a course that “involves the supervised practical application of previously studied theory.” This collection of courses is designed to give students concise, hands-on experience with both (a) the contextual elements of data work that are best done by them, as humans with domain expertise; and (b) collaborating with AI to perform a full range of data tasks that they might not otherwise be familiar with. The purpose is not so much to teach them the theory and practice of advanced statistical procedures – a 1-credit hour course simply does not have the time for that – but, rather, how they can work productively with AI to help them with new tasks in an effective and efficient way.
These 3 special topics courses all focus on applied data analysis, or data work that is designed to give actionable insights to identifiable stakeholders in support of their practical goals (as opposed to theory-building research that is generally addressed to a field as a whole). Each course is designed as a 1 credit hour, online course, with instruction delivered online through videos and downloadable materials. (These materials would be embedded in Canvas but would also be generally available as OER on YouTube, datalab.cc/practicum, and elsewhere with a Creative Commons attribution-only, or CC BY, license.) In addition to the online materials, I would be available to meet with enrolled UVU students for in-person office hours or a group meeting 1 hour per week (Mondays at 2:00 pm) or possibly at other times through Teams.
The human side of AI-first data work
These courses are designed to help students with “AI-first data analysis,” or data work in which the majority of the technical work is conducted by AI, while people (like you) are responsible for all of the non-coding, durable skills that gives data projects value:
Exploring goals with stakeholders
Framing actionable questions
Engaging with the AI to choose appropriate analyses
Auditing the analytical process
Exploring and interpreting results
Deriving actionable insights
Communicating recommendations with stakeholders
Tools to be used
These courses will use the following tools:
The Python programming language, which is currently the most common tool in applied data science and the foundational language of AI
Google Colab, which is an online programming environment that is based on the user-friendly Jupyter Notebooks. Colab has both a free tier for the general public and, for students, 12 months of free access to Google AI Pro. Colab is integrated into Google Drive accounts and greatly facilitates the management of Python environments. In addition, it is accessible from any web-connected device, such as Chromebooks, university computers, and even mobile devices like tablets and smartphones, which promotes equal access with few resource restrictions. Colab also has excellent capabilities for sharing work with collaborators, instructors, and stakeholders.
Google Gemini AI, a generative AI model that is built into Colab (and many other Google services). Gemini can create notebook sections and write code in response to prompts from students, as well as collaborate in the interpretation and application of results.
Qualtrics, a proprietary online form/survey application that is used by organizations globally. UVU students have free access while enrolled.
Google Forms is an online form/survey application that has the major advantages of (a) being built into Google Drive; and (b) being free, so it is affordable by students and stakeholders when they no longer have access to paid Qualtrics accounts.
Integrated Google tools such as Google Drive, Google Sheets, and Google Slides, all of which are free and familiar to students and stakeholders of all kinds.
(Note: No proprietary or sensitive data will be used in the first course in this sequence. Instead, all datasets used will be either sample datasets that are included with Python or simulated datasets that I create to illustrate topics like data cleaning, missing data, or joining datasets.)
Reminder: Discussions are brief and assume independent, AI-guided exploration
These courses are sketches, not detailed paintings.
That is, a broad range of topics are briefly covered in the weekly schedules for these courses. Please remember, though, that these classes are each 1 credit hour, which assumes 1 hour of instruction per week, with 2-3 hours of independent work outside of class per week. As such, instruction is NOT comprehensive. The intention is, rather, to do the following:
Let students know that a topic exists, such as the social contract for research or confidence intervals for effect sizes
Give a few minutes of explanation of what it is, why it matters, and when and how it would be used
Provide a short demonstration of how students can prompt the AI for the task, such as choosing among several procedures, preparing the data for the analysis, choosing options for analysis, troubleshooting, reviewing output, preparing graphics, and discussing possible interpretations.
This means that topics that might merit an hour (or week or semester) of their own will, instead, be briefly discussed in 5-15 minutes. This is not because the topic is not important or for the lack of things that I, as the instructor, could say about it. Rather, it is because these courses are designed, above all, to foster the students’ inquisitive collaboration with AI to help with a wide range of technical tasks, relying on the potential for dramatically increased efficiency in AI-guided work. (This approach is not based merely on “AI is magic” hyperbole but on my own experience in academic and professional work.)
What this means is that, in addition to fully using the resources provided in the class and the instructor’s time, students will assume a substantial responsibility for working productively with AI on their own. As they say in the programming world, “It’s not a bug, it’s a feature.”
Data Practicum 1: Actionable Insights with AI (Spring of 2027)
UVU: PSY 4750R-X01: Current Topics PSY Insights AI (1 CR, online, CRN 42254)
This course introduces an end-to-end, AI-assisted workflow for turning raw data into clear, actionable recommendations. Designed as a practical bridge from introductory statistics, students use Google Colab, Python, and Gemini to formulate stakeholder questions, design surveys in Google Forms and Qualtrics, clean and score psychometric data, and run foundational analyses. To safeguard privacy and establish ethical habits early, all coursework uses simulated datasets and standard open-source libraries rather than real client data. Emphasis is placed on practical data ethics – including FERPA, HIPAA, and client privacy – as well as auditing AI decisions, documenting analysis pipelines with verification notes, clearly communicating insights and limitations, estimating return-on-investment (ROI), and preparing presentation-ready reports and visualizations.
Weekly schedule
Week 1: Intro to applied analysis
Applied analysis vs theory-building research. Set up Colab and Gemini. Demonstrate a quick, end-to-end project on simulated data.
Week 2: Actionable insights, ROI, and ethics
Asking questions that can be answered with data. Actionable insights and ROI. The client contract. Ethics. Intuitive priors as decision aids.
Week 3: Python, Colab, and simulated data
Python and Colab basics. File structure, comments, data dictionaries. Privacy and AI: HIPAA, FERPA, CPRA, GDPR, etc. Methods for verifying AI output.
Week 4: Google Forms and Qualitrics
Designing forms in Google Forms and Qualtrics. Honoring the “social contract” in research. Designing forms that allow flexibility but are easy to analyze.
Week 5: Importing "tidy data"
Defining “tidy data.” Importing and tidying from Qualtrics and Google Forms with Python code. Naming variables. Simple version control.
Week 6: Data cleaning
Cleaning data with reproducible Python code. Data types, missing values, duplicates, outliers and unusual values. Documenting work.
Week 7: Data coding and scoring
Scoring. Reverse-scored items, scale scores, scale reliability (alpha, etc.), recoding categories, creating scale scores.
Week 8: Graphs for insight
Graphs as analyst tools for exploratory data analysis (EDA) as opposed to graphs for presentation. Univariate, bivariate, multivariate graphs for insight.
Week 9: Estimating means and effect sizes
Estimating means, percentages, ratios, and effect sizes with confidence intervals. Emphasis on justifiable interpretations for frequentist methods.
Week 10: Comparing groups with t-tests and ANOVA
Comparing group means with t-tests and ANOVA. Choosing tests, checking assumptions, exploring follow-ups.
Week 11: Correlation and regression
Correlations, correlation matrices, and simple versions of multiple regression. Begin work on projects.
Week 12: Presenting results
Best practices for presenting results. Lead with actionable insights, put labels and todo items on each graph. Address limits, include appendix of analyses.
Week 13: Documenting and archiving work
Documenting. Random seeds, package versions, AI-use statements, methods to save and share data while preserving privacy and facilitating future work.
Week 14: Overview of advanced topics
An overview of topics for Data Practicum 2 and 3, and other resources. Develop greater confidence and independence in collaborating with AI on new analyses.
Week 15: Project presentations
Project presentations.
Data Practicum 2: Open Data & Decision Support (Fall of 2027)
Building on foundational AI-assisted workflows, this course engages the messy reality of open and public data to solve real community and organizational problems. Students locate, evaluate, and merge open datasets such as Google Trends and municipal records, addressing issues of data provenance, joining datasets, and missingness before conducting multivariable analyses beyond introductory statistics. Topics to be briefly covered include multiple regression, model validation (holdouts and cross-validation), Bayesian perspectives, and an introduction to advanced methods like text mining and multilevel models. The semester culminates in an archived, reproducible analysis and an executive decision-support report delivered directly to a nonprofit stakeholder.
Weekly schedule
Week 1: Intro to open data and applied work
Review the workflow template and Practicum 1 content. Choose project questions.
Week 2: Open data
Finding and judging open data. Provenance, what each variable measures, and a stakeholder profile built from public material.
Week 3: Google Trends
Google Trends in depth. Relative scale, sampling, saved downloads, and the challenges of time series data.
Week 4: Combining data
Combining sources. Merges, keys, and silent row loss. Matching towns across datasets.
Week 5: Planning analyses
Planning analyses before looking at results. Precision and sample size, and a short written analysis plan.
Week 6: Regression models
Multiple regression, including block-entry regression, LASSO, and hierarchical models. Effect sizes.
Week 7: Interpreting models
Exploring multiple potential causes. Confounding, the ecological fallacy, and careful causal language.
Week 8: Validating models
Validation. Holdout, cross-validation, and seeds, taught as "will this hold up on new data?"
Week 9: In-depth comparisons
Group comparisons in depth. Strategies to reduce false positive errors. Post-hoc tests and multiple comparisons.
Week 10: Bayesian methods
The Bayesian lens. Priors and sensitivity to priors, with the costs named and compared with intervals.
Week 11: Data mining overview
Overview of LinkedIn Learning data mining courses that are free to UVU students covering topics such as clustering, classification, or text mining.
Week 12: Research synthesis and meta-analysis
Synthesis and multilevel models. Theoretical and practical methods for dealing with nested data structures.
Week 13: Creating reports
Drafting the stakeholder report. Graphs first.
Week 14: Standards for reporting
Peer review, revision, archiving (OSF or GitHub), and APA-style reporting of methods and AI use.
Week 15: Project presentations
Final presentations. Send the reports to the nonprofits.
Data Practicum 3: Independent Consulting (TBD)
In this capstone experience, students operate as independent data consultants, taking complete ownership of an external client engagement from project scoping to final delivery. Working directly with community partners, businesses, or research labs, students define the analytical problem, navigate data constraints, choose appropriate methods, and produce a polished, professional deliverable tailored to the client's needs. With a focus on professional autonomy, reproducible archiving, and persuasive data storytelling, this course prepares students to confidently lead applied analytical projects in industry, non-profit, or graduate settings.
Weekly schedule
Topics and schedule to be determined in consultation with students.
