Module 01: AI in 2026 — Landscape, Opportunities & Leveraging AI in College

Learning Objectives

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

1. Map the AI stack — Describe layers (data, models, apps, governance) and relate them to how trust and errors actually show up in your work.
2. Contrast paradigms — Explain when generative AI, classical ML, or rules-based automation fits—and when mixing them without documentation hurts your methods section.
3. Critique claims with evidence — Separate benchmark hype from “would this work on my task?”
4. Design a defensible workflow — Align tools with honor codes, data boundaries, and disclosure you could show an advisor.
5. Connect literacy to careers — See orchestration, evaluation, and domain depth as the real stack—not “prompting only.”

Deep Concept Explanation: Academic Context, Research Methodologies, and Strategic Framing

1.1 The 2026 AI stack as a socio-technical system

Think of AI like a concert: you see the singer (the chat UI), but sound engineers, venues, ticket apps, and city permits all matter. In college work, trace which layer produced each output—retrieval, base model, safety filters, or your own prompt. Research in HCI and STS treats this as a socio-technical system: tech and institutions co-evolve.

1.2 Generative AI, classical ML, and rules

1.3 Epistemology: hypothesis vs. oracle

Treat model output as hypothesis-generating unless you ground it in sources, tools, or data you control. Split divergent brainstorming from convergent verification—same instinct as double-checking a lab measurement.

1.4 Research workflows with AI

Literature — AI can suggest keywords or table shells; you open PDFs and verify DOIs. Code/data — AI can draft snippets; you own seeds, versions, and leakage checks. Qualitative — disclose machine-assisted coding; protect human-subjects data.

1.5 Innovation without solutionism

Strong projects start from validated pain (“problem pull”), not from “we have a model, let’s bolt it on.” That mirrors lean hypotheses and responsible innovation: anticipate harms, include affected voices.

1.6 Policy, integrity, and careers

Sylla vary—ask early when unclear. Employers care about workflow design (retrieval, eval gates, logging) plus domain depth—not vibe-based prompting.

Code and Computational Examples

Key Example: A minimal AI use log helps you disclose assistance honestly and builds a habit your capstone can reuse.

[object Object],
,[object Object], dataclasses ,[object Object], dataclass, asdict
,[object Object], datetime ,[object Object], date
,[object Object], pathlib ,[object Object], Path
,[object Object], csv

,[object Object],
,[object Object], ,[object Object],:
    entry_date: ,[object Object],
    course_or_project: ,[object Object],
    task_description: ,[object Object],
    tool_name: ,[object Object],
    prompt_summary: ,[object Object],
    human_verification: ,[object Object],
    disclosure_required: ,[object Object],

LOG_PATH = Path(,[object Object],)

,[object Object], ,[object Object],(,[object Object],) -> ,[object Object],:
    new_file = ,[object Object], LOG_PATH.exists()
    ,[object Object], LOG_PATH.,[object Object],(,[object Object],, newline=,[object Object],, encoding=,[object Object],) ,[object Object], f:
        w = csv.DictWriter(f, fieldnames=,[object Object],(asdict(entry).keys()))
        ,[object Object], new_file:
            w.writeheader()
        w.writerow(asdict(entry))

,[object Object], __name__ == ,[object Object],:
    append_entry(AIUseEntry(
        entry_date=,[object Object],(date.today()),
        course_or_project=,[object Object],,
        task_description=,[object Object],,
        tool_name=,[object Object],,
        prompt_summary=,[object Object],,
        human_verification=,[object Object],,
        disclosure_required=,[object Object],,
    ))

Instead of extra code blocks, use a stack map in prose: list Data → Model → UI → Evaluation → Governance and one failure mode + mitigation per row. JSON/YAML experiment cards work the same way on paper.

Practice Exercises

1. Fit matrix — Eight tasks from your major; rate AI fit (high/medium/low) plus one verification step each.
2. Policy close-read — Five must-dos, three no-gos, three gray areas + a draft email to an instructor for one gray area.
3. Headline audit — Five articles: claim type, evidence type, conflict of interest—note the weakest.

Mini-Project: Personal AI Charter & Research Integrity Plan

Scope: 2–3 pages: permitted/prohibited uses by context, verification checklist, data you will never paste without approval, stack map, incident playbook (e.g., fake citation, teammate pastes sensitive data), short career narrative.

Discuss with a friend: Trade charters and mark one blind spot each.

Key Takeaways

Academic integrity and authorship

Your name means you stand behind the reasoning. AI assists; it does not assume liability. Disclose stages (ideation, code, polish) when required.

Conceptual anchors

• The landscape is a stack—not “the chatbot.”
• Fluency ≠ truth; verify or label uncertainty.
• Problem-pull beats tech for its own sake.

References and Cross-Modules

Modules 03, 09, 10–12 extend this into writing, evaluation, entrepreneurship, industry, and capstone. Refresh your personal charter after major model or policy shifts.