AI & Code

Claude Code, Codex, and similar coding agents become useful once thesis work turns file-based. At that stage you have source folders, metadata sheets, scripts, and outputs that need checking against the originals. Methods documentation belongs in the same workflow. The point is inspection and rerunning. Judgment and final interpretive claims stay with you.

Choose a Workflow

Starter Prompts

Treat these as starting points. Replace the bracketed text with your actual project details.

I am writing a thesis on [topic]. Help me set up a transparent research folder for sources, metadata, notes, scripts, outputs, and documentation. Suggest a folder structure, file naming convention, and metadata columns. Do not analyze the sources yet. Focus on organization and reproducibility.

I have [number/type] source files about [topic] from [source/database] covering [date range]. Help me design a corpus-building workflow. I need inclusion/exclusion rules, metadata fields, file organization, quality checks, and a short methods note explaining the workflow. Ask clarifying questions before proposing scripts.

Inspect this script as a research workflow. Treat the code as one part of that workflow. Explain what each step does, identify assumptions, list possible failure points, and suggest validation checks. Do not change the analytical logic unless you explain why.

Help me draft a transparent methods note describing how AI/code tools supported my workflow. Include what the tool helped with, what I checked manually, what remained my responsibility, and where scripts or prompts are stored. Keep the wording factual and concise.

Responsible Use Checklist

Before you rely on any AI- or code-assisted workflow, you should be able to answer these questions.

• Can you explain what the tool did and why that step was appropriate?
• Are prompts, scripts, outputs, and manual corrections stored in a place you can revisit?
• Did you verify a sample of outputs against the original sources?
• Are the corpus boundaries documented, along with any exclusions and transformations applied along the way?
• Is the final interpretation yours, grounded in evidence you checked?
• Have you discussed the planned use with your supervisor and followed the relevant Ethics & AI policy on permission and disclosure (and the broader academic-integrity expectations)?

Good Division of Labor

The standard is simple. If the work shapes the argument, you have to understand and defend it. For repetitive or procedural tasks, the tool can help. Verification is still on you.

Where To Go Next