Project 1 — Read Like a Scientist

“It is the glory of God to conceal things, but the glory of kings is to search things out.” — Proverbs 25:2 (ESV)

Chapter: 1 — What Is Research, Really? Due: End of Week 1 Submit: A link to your research-portfolio Git repository (public or instructor-invited) containing domain.txt, three structural analyses, a verified references.bib, your built paper template under paper/, REFLECTION.txt, and (Hard tier) MEMO.txt. See Appendix A for portfolio + Overleaf setup and submission mechanics, and Appendix B for the literature toolkit. Allowed tools: Any literature database (Google Scholar, ACM DL — fully open access as of Jan 1, 2026 — IEEE Xplore, arXiv, Semantic Scholar, Connected Papers), Zotero, Git/GitHub, Overleaf, python3, and an AI assistant for finding and summarizing only. AI policy (Phase 1): AI may help you find and summarize papers; it may not supply a citation you do not personally verify, and it may not write your analysis or reflection. Every AI-touched step carries an honest AI usage note. The human owns every claim and every citation. (No agent-log.txt yet — that begins in Phase 2; AI-disclosure rules live in Appendix C.)

The Setup

You are at the start of your master’s practicum, and a thesis-shaped year is in front of you. Your advisor — a busy person who has read ten thousand papers and started a hundred projects — asks you one question over coffee: “What are you going to work on, and do you actually understand the field you’re walking into?”

You cannot answer that honestly yet. So this week you earn the right to. You will commit to a domain, read three real papers in it the way a scientist reads — fast, structured, skeptical — and start the two artifacts that turn this course from a class into a submission: a version-controlled research portfolio and a paper in a real venue template. When your advisor asks again next week, you will hand them three analyzed papers, a sharp question, and a repo that proves you have already started. That is what “ready” looks like. The work below gets you there.

Setup (the starter)

This chapter ships starter and reference files in code/. Copy them into your portfolio repo and make them yours:

• code/paper-anatomy.txt — the section-by-section reading map and the three-pass method. Your reading template.
• code/structural-analysis-template.txt — one form per paper. You fill out three.
• code/paper.tex — your ACM acmart skeleton with sections stubbed and AI-disclosure placed correctly. Drop it into Overleaf.
• code/verify_citation.py — checks a citation against Semantic Scholar so you verify by eye. Run it on every reference.

Learning Targets

You will demonstrate that you can:

• Choose and justify a CS domain and frame a single falsifiable, AI-meets-domain research question.
• Read research efficiently with the three-pass method and produce a structural analysis of three real papers — naming each section’s job and rating the four pillars with evidence.
• Verify every citation against a real index before it enters your portfolio, and report on AI-supplied citation reliability honestly.
• Stand up the research infrastructure — a Git portfolio and an ACM/IEEE paper template — that the next fifteen weeks build on.

Normal Tier

Goal: Commit to a domain, read and structurally analyze three recent papers, verify every citation, and stand up your portfolio and paper template.

Required features

1. domain.txt — Your chosen domain (from the §1.6 menu) and one falsifiable, AI-meets-domain question in the form “Does [AI method] improve [outcome] in [domain], measured by [metric]?” Plus two sentences on why this domain.
2. Three structural analyses — One completed code/structural-analysis-template.txt per paper (three total), for recent conference papers in your domain. Each must include the actual section map, the research-type classification, and the four-pillar ratings.
3. references.bib — A BibTeX file with all three papers, each verified against dblp / Semantic Scholar / the publisher. Note in a comment how you verified each.
4. Built paper template — code/paper.tex (or an IEEEtran equivalent) compiling in Overleaf or locally, with your title and a stubbed abstract, committed under paper/. Record the literal template version string.
5. Portfolio repo — Initialized with the structure from Rep 10, real commit history, a README.txt naming your domain and question.
6. REFLECTION.txt — 500–800 words: what research is (in your own words, not a definition copied), what kind of research you intend to do and why, and the gap (from a real Future Work section) you might one day occupy. Include the AI usage note.

Normal-tier rubric (out of 100)

Medium Tier (+up to 25% extra credit)

Go deeper than reading — start mapping the field.

1. Snowball one paper. Take your most central paper and use Connected Papers (or Research Rabbit / Semantic Scholar’s citation graph) to find five neighbors — three it cites (backward) and two that cite it (forward). Verify each, add to references.bib, and write one line on how each relates.
2. Beck’s four-sentence abstract. For your intended project, draft an abstract using Kent Beck’s four moves (problem / why it’s a problem / the contribution / the implication) — even though you have no results yet. This is a plan, and it goes in your paper’s abstract stub.
3. The CARS intro. Write the three-move opening paragraph for your intro (Swales: establish the territory → establish the niche/gap → occupy it), with the gap backed by a cited absence from one of your papers.

Hard Tier (+up to 25% additional extra credit)

This is the judgment no tool can make for you, and it is graded as such.

Write MEMO.txt (one page, addressed to “my advisor”) that takes a defensible position: which one of the four pillars — novelty, rigor, reproducibility, or significance — is the real bottleneck for progress in your chosen domain right now, and what that implies for where you should aim your project. You must:

• Name the pillar and defend the choice with evidence from the papers you read (e.g. “every paper in my set reports a single seed — rigor, specifically variance, is the field’s weak point”).
• State what this implies for your project: if reproducibility is the bottleneck, a rigorous replication may be more valuable than a flashy new method.
• Tie it to the spine rule — the human stays in the loop where the judgment lives — and name one integrity risk (fabricated data, hallucinated citation, single-run claim) your project will explicitly guard against.

A memo that says “all four matter equally” fails. The grade is in the discrimination: showing you can read a field, find its weakest joint, and aim at it.

Submission

Push everything to your portfolio repo:

research-portfolio/
  README.txt
  domain.txt
  lit/
    paper-1-analysis.txt
    paper-2-analysis.txt
    paper-3-analysis.txt
  paper/
    paper.tex
    references.bib
  REFLECTION.docx
  MEMO.docx            # Hard tier only
  reps.txt

Submit the repo link per Appendix A. Use placeholders for anything sensitive; never commit a real credential or a non-public dataset.

Hints (Read Before You Begin)

• Choose the domain first and do not look back. Switching in Week 5 means redoing your literature review. Pick something you can stay curious about for sixteen weeks.
• Recent and conference, not ancient and journal. You want the current shape of the field. ACM DL went fully open access on Jan 1, 2026 (free Basic = full corpus); arXiv for preprints; Semantic Scholar for TLDRs and the citation graph.
• Verify before you read deeply. A fabricated citation wastes the hour you spend reading the paper that does not exist. Run code/verify_citation.py on the title first.
• Back every pillar rating with a line. “Rigor 4/5” with no evidence is worth nothing. “Rigor 4/5 — five seeds, 95% CIs, Table 3” is analysis.
• An empty template is a map, not a waste. Start the paper now. You will thank yourself in Week 14.

What Mastery Looks Like (Beyond the Rubric)

A mastered submission is one where I can read your three analyses and your MEMO.txt and tell that you read these papers — that you saw which one reports a single seed and which releases its code, that your question could genuinely be wrong, and that you can name the one weak joint in your field and aim at it. The reading is table stakes. The judgment — the verified citations, the evidence-backed ratings, the defended position on what your field actually needs — is the project.

Coach’s Note — The temptation this week is to skim three abstracts, let a chatbot summarize them, and call it analysis. That produces a report I have read a thousand times and learned nothing from. The honest, slower path — three real Pass-2 reads, every citation checked by your own hand — is the path that makes you a researcher instead of a summarizer. Your credibility for the next fifteen weeks starts here.

When You’re Done (a short checklist)

• domain.txt names one domain and one falsifiable AI-meets-domain question.
• Three structural analyses, each with a real section map and four-pillar ratings.
• Every rating points at a specific line, table, or figure.
• Every citation in references.bib is verified; the method is noted.
• The paper template builds; title + stubbed abstract; version string recorded.
• The portfolio repo has structure and a real commit history.
• REFLECTION.txt is in your own words and includes the AI usage note.
• (Hard) MEMO.txt defends one pillar as the field’s bottleneck.
• No fabricated citation, no copied definition, no real secrets committed.

A theological footnote. “The glory of kings is to search things out” (Proverbs 25:2, ESV). This project is the smallest rehearsal of that glory: not making knowledge, but searching out what is already there and telling the truth about it. The chatbot will offer to hand you a tidy summary and a list of references — and some of those references will not exist. To accept them unread is not efficiency; it is a small lie about the world, and it forfeits the dignity the verse names. To verify every citation, to back every rating with what the paper actually says, to admit which pillar your field is weakest on — that is what it means to search a matter out faithfully. You are handing the next researcher knowledge they can stand on. Start as you mean to go on.

See you next week.