Preface

This volume is a one-week workshop for faculty researchers. We have in mind the reader who runs a grant-funded project, publishes under their own name, and usually teaches and supervises trainees as well, and who wants AI tools to make that entire workload faster. We assume an active researcher, not a beginner to research itself. Where a general-audience book would teach prompting from nothing, this one teaches specific techniques for getting real research work done faster (compiling a bibliography, scaffolding a project, drafting an outline), together with the judgment that keeps that speed from costing the researcher a retraction, a misconduct finding, or a withdrawn grant. The goal is a working, disciplined practice, not a tour of features.

The design constraint is deliberate: five lectures, ten hours of homework, and the reader finishes the week doing the same research work in substantially less time, while knowing precisely where the tools that bought that speed are reliable and where they are not.

Efficiency first, discipline in service of it

The primary reason to take this workshop is speed, not safety. A faculty researcher juggling grant deadlines, a teaching load, and a lab full of trainees does not have a spare afternoon to hand-assemble a bibliography one reference at a time, or to scaffold a new project’s directory structure from memory, or to stare at a blank page until an argument’s shape appears. Each day of this workshop presents a specific technique for doing that work faster: batch-compiling a working bibliography from a database of record instead of one-by-one chat prompts (Day 2); using a coding assistant to bootstrap a reproducible project skeleton instead of copying one by hand (Day 3); and building a structured outline through a short clarifying dialog instead of drafting blind (Day 4).

Unfortunately, none of that speed is worth much if the work has to be redone. That is the entire reason this book also treats the recognition of hallucination and fabricated citations, the review of AI-generated code, the protection of confidentiality, and the honest disclosure of AI use. Safety is not the point of the week. The point is that a faculty researcher who works fast on a journal’s, a funder’s, or an IRB’s terms, and who then has to retract a citation, redo an analysis, or explain an undisclosed AI contribution, has not actually saved any time at all. We may consider three frictions that make this concrete:

  • A literature-review assistant can fabricate plausible citations that do not exist, and the fabrication is invisible unless every citation is checked against a primary source, which is exactly why Day 2 teaches querying a database of record directly rather than trusting a chatbot’s free-text citation list in the first place.
  • A coding assistant can produce code that runs and looks correct while silently mishandling edge cases, missing data, or unit conversions, which is why Day 3 pairs the scaffolding technique with a review habit for anything the assistant writes beyond the skeleton itself.
  • A writing assistant can smooth prose into a voice that is not the author’s, and journals increasingly require explicit disclosure of where AI assistance was used, which is why Day 4’s outlining technique keeps the AI’s role bounded to structure while the argument stays the author’s own.

The verification and disclosure habits in this book are the guardrails that make the efficiency gains sustainable across an entire career, not a tax on using AI tools at all.

What this book covers

We proceed over five days, one chapter each. We begin in Day 1 with the foundations, turn in Days 2 through 4 to the three efficiency techniques that give the week its payoff, and close in Day 5 with the provenance and disclosure practice that makes the whole arrangement defensible:

  1. Foundations: how LLMs work and where they fit in research. What a large language model does and does not do; tokens, context windows, and hallucination as a structural property, not a bug; choosing a tool for a task; the shape of a good prompt.
  2. Literature review and synthesis with AI tools. Compiling a working bibliography efficiently, from a database of record rather than one-by-one chat prompts; AI-assisted search and summarization; verifying citations against primary sources; synthesizing across papers without flattening disagreement; the specific risk of fabricated references and how to catch it.
  3. AI-assisted coding and data analysis. Scaffolding a reproducible project skeleton with a coding assistant such as Claude Code or GitHub Copilot; working with a coding assistant more generally; reviewing AI-generated code for correctness, not just plausibility; using AI to accelerate data cleaning and exploratory analysis while keeping a human in the loop for inference.
  4. Writing, editing, and citation management with AI. Generating a structured outline through chatbot dialog before drafting any prose; drafting and revising with AI assistance while preserving the author’s argument and voice; reference-manager integration; what journals currently require for AI-use disclosure.
  5. Reproducibility, provenance, and responsible use in publication. Recording what AI tools did in a project; version-controlling prompts and AI-assisted artifacts alongside code; disclosure norms and venue policies; a checklist for auditing a finished manuscript for undisclosed or unverified AI contributions.

An appendix extends Day 5’s discipline into three roles specific to faculty life: grant writing under funder-specific AI rules, setting AI policy for one’s own students, and resolving AI-related authorship and credit questions with one’s own trainees.

Each chapter is approximately 1 hour of reading and worked examples plus 2 hours of homework problems with worked solutions provided.

What this book does not cover

The book deliberately omits, and points elsewhere for, material that surrounds AI-assisted research but is not the subject of this workshop itself:

  • Learning R, Python, or statistics from scratch. We assume the reader arrives already doing research in some discipline. See R for Biostatistics.
  • Version control with Git and GitHub. Day 5 uses Git to track AI-assisted work, but does not teach it. See Git and GitHub for Biostatistics.
  • Containerized and dependency-pinned reproducibility in depth (validated pipelines, multi-container orchestration, renv at the scale of a large team). Day 3 covers using an AI assistant to bootstrap a basic renv/Docker project skeleton; the deeper practice sits in the Biostatistics Practicum and Reproducible Research for the Health Sciences.
  • Building or fine-tuning language models. We use existing tools as a researcher would, not as a model developer would.

The workshop is intentionally narrow. The reader finishes Day 5 able to use AI tools across a real research project while keeping the work verifiable, attributable, and reproducible.

How this book is meant to be used

The five-day cadence is the spine. A reader starting the week using AI tools informally can finish the week with a disciplined, disclosed, and reproducible practice. The cadence assumes roughly 3 hours of focused work per day:

  • Hour 1: Read the chapter, running every example against an AI tool of your choice as you reach it. The judgment this book teaches is learned by checking real outputs, not by reading about them.
  • Hours 2-3: Complete the homework problems. Solutions are at the end of each chapter; check yourself only after attempting each problem.

The examples in this book are tool-agnostic where possible and note tool-specific behavior where it matters.

Prerequisites

The book assumes:

  • Active involvement in an academic research project, in any discipline.
  • Access to at least one general-purpose AI assistant (for example Claude or ChatGPT) and, for Day 3, a coding assistant.
  • Basic familiarity with the command line: opening a terminal, changing directories with cd, listing files with ls.

Prior completion of R for Biostatistics and Git and GitHub for Biostatistics is helpful for Day 3 and Day 5, since the worked examples assume a version-controlled project, but neither is strictly required; the AI-use judgment this book teaches is independent of the specific tools used to manage the underlying project.

Acknowledgments

This workshop consolidates patterns developed while integrating AI tools into ongoing research and teaching work. The documentation for the AI tools in current use, and the emerging literature on AI-use disclosure in scholarly publication, provide the technical scaffolding; the workshop pedagogy reflects experience with what a working researcher actually needs in order to use these tools well, as distinct from the full capabilities of any one tool.