Module 03: Writing with AI — Essays, Reports & Presentations with Academic Integrity

Learning Objectives

By the end of this module, you will be able to:

1. Stage invention → structure → draft → revision → editing with explicit human checkpoints and syllabus alignment.
2. Execute citation workflows: open every source, stable IDs, zero ghost references.
3. Design slides and reports where evidence leads—one main claim per slide, sourced visuals.
4. Document disclosure, version history, and permitted vs prohibited substitution.
5. Evaluate co-authored text for generic “AI voice,” fabricated specificity, and missing provenance.

Deep Concept Explanation: Academic Context, Research Methodologies, and Communication Rigor

3.1 Writing as knowledge work

Writing fixes claims so they can be criticized. LLMs optimize local coherence, not your thesis or fair treatment of sources. You endorse truth, fairness, and ethics.

3.2 Where AI fits

3.3 Citation integrity

1. Open every cited source. 2. Prefer DOI/library links. 3. Never fabricate page numbers. 4. Treat AI reading lists as search hypotheses.

3.4 Disclosure

State what tool, when in the process, how verified. Process appendices often beat vague footnotes.

3.5 Presentations and group work

One claim per slide; source lines on figures; version control with named edits; align team AI norms before drafting.

3.6 Upload ethics

Avoid pasting unreleased theses, employer IP, identifiable human-subjects text, or full paywalled PDFs where forbidden.

Code and Computational Examples

Key Example: Before a cite enters your bibliography, require a DOI or HTTPS URL—this dataclass pattern keeps the rule explicit.

[object Object], dataclasses ,[object Object], dataclass
,[object Object], typing ,[object Object], ,[object Object],
,[object Object], re

,[object Object],
,[object Object], ,[object Object],:
    title: ,[object Object],
    year: ,[object Object],[,[object Object],]
    doi: ,[object Object],[,[object Object],]
    url: ,[object Object],[,[object Object],]

DOI_RE = re.,[object Object],(,[object Object],, re.I)

,[object Object], ,[object Object],(,[object Object],) -> ,[object Object],:
    ,[object Object], stub.doi:
        ,[object Object], ,[object Object],
    ,[object Object], stub.url ,[object Object], stub.url.startswith(,[object Object],):
        ,[object Object], ,[object Object],
    ,[object Object], ,[object Object],

Without code: Use a process appendix table (Stage | Human work | Tool | Verification) and a claims traceability list mapping each empirical claim to R1…Rn or a dataset path.

Practice Exercises

1. Voice calibration — 300 words solo, then neutral rewrite; restore two irreplaceable phrases.
2. Citation rescue — Delete or support every sentence in an AI “background” paragraph.
3. False citation hunt — Five model-suggested papers; log hit rate.
4. Integrity scenarios — Exam, take-home, lab, thesis: allowed / prohibited / disclosure for your school.
5. Collaborative log — Group table: who wrote what + AI use per member.

Mini-Project: Policy-Compliant Essay Package

Essay (1200–1500 words) + process appendix + six sources (four scholarly primary) + citation verification cover sheet + optional before/after polish diff for one section.

Key Takeaways

Epistemic hygiene

Fluency ≠ accuracy; citation is evidence routing.

Presentation specifics

Speaker notes carry depth; alt text and licenses on visuals.

References and Cross-Modules

Connects to Module 09 (evaluation) and Module 11 (team docs). Disclose retrieval limits for RAG-based assistants in methods.