{"course":{"slug":"ai-for-students","title":"AI for Students","url":"https://addaly.com/learn/ai-for-students","level":"beginner","track":"work"},"lesson":{"slug":"reading-a-paper-fast","title":"Reading a paper without reading all of it","url":"https://addaly.com/learn/ai-for-students/reading-a-paper-fast","position":41,"of":73,"minutes":8,"answer":"Read in the order abstract, figures, aim, discussion, limitations, and check any claim that matters against the results section rather than the abstract, because abstracts systematically overstate.","keyIdea":"Read in the order abstract, figures, aim, discussion, limitations, and check any claim that matters against the results section rather than the abstract, because abstracts systematically overstate."},"body":{"format":"markdown","text":"## Nobody reads papers front to back\n\nResearchers do not start at the first word and finish at the last. They triage. For an undergraduate reading twelve papers for an essay, the difference between triaging and not is the difference between finishing and not.\n\nThe order that works:\n\n1. **Title and abstract.** What did they do, what did they find. Decide whether to continue. Most papers stop here, and that is a successful outcome, not a failure.\n2. **Figures and tables, with their captions.** This is the surprising one. The figures usually contain the actual result, and captions in a good paper are written to be readable alone. You can often learn what a paper found in ninety seconds this way.\n3. **The last paragraph of the introduction.** Where the specific aim and hypothesis live.\n4. **The first and last paragraphs of the discussion.** What they think it means, and what they concede.\n5. **The limitations section.** Read this properly. It is where the honest authors tell you what the paper cannot support, and it is exactly what you need in order to cite it accurately.\n6. **Methods**, only if the result matters to you. Then read it slowly, because this is where a study is strong or weak.\n7. **Results text**, to check the numbers match what the abstract claimed. They sometimes do not.\n\n## Where the model helps and where it must not\n\nThe safe pattern: read the paper yourself in that order, and use the model on the parts you cannot follow, grounded on the pasted text.\n\n> Figure: The order researchers actually read in. Steps: 1. Abstract (What they did and what they found. Decide whether to go on.); 2. Figures and captions (Usually the actual result. Good captions are written to be read alone.); 3. The aim (Last paragraph of the introduction, where the hypothesis lives.); 4. Discussion and limits (First and last paragraphs, then the limitations, where honest authors say what it cannot support.); 5. Results, to check (Does the number match what the abstract claimed? Sometimes it does not.). Methods come last, and only if the result matters to you. The step almost nobody does is the final one: abstract conclusions have a documented tendency to be stronger than the results section supports, so a claim carrying weight in your essay gets checked against the results, which takes a minute. Stopping after the abstract is a successful outcome, not a failure.\n\n> Here is the methods section. Explain what a mixed-effects model is doing here and why they would have chosen it over a repeated-measures ANOVA. Use only what is in the text; say what is missing rather than filling it in.\n\nThat is a good use. It is a comprehension aid on material in front of you.\n\nThe unsafe pattern is asking for a summary and citing the summary. Two things go wrong. Summaries drop conditions and hedges (the *in older adults only*, the *when baseline was controlled*) and those qualifications are frequently the entire finding. And a summary you did not check gives you no page number and no sentence, so you cannot cite honestly.\n\n## The questions to ask of any study\n\nWhatever the field:\n\n- **How many participants or observations, and how were they selected?** A study of 24 undergraduates supports different claims from one of 24,000 adults.\n- **Compared with what?** No control group means no causal claim. A comparison with nothing is not a comparison.\n- **Was anything randomised?** This is the difference between *associated with* and *caused*.\n- **What exactly was measured?** *Wellbeing* measured how? Self-report on one item is not the same as a validated scale.\n- **How many analyses were run?** Twenty tests produce a significant result by chance.\n- **Who funded it, and who benefits from the finding?** Not disqualifying; relevant.\n- **Is this the first report or a replication?** First reports are much more often wrong than the confidence of their abstracts suggests.\n\nA model can be asked to apply this list to a paper you have pasted, and that is a good use: it is applying a checklist to supplied text rather than recalling anything. Verify the answers against the paper.\n\n## Abstract inflation\n\nWorth knowing as a specific hazard: abstracts systematically overstate. They are written to be read and cited, they omit the qualifications, and there is a documented tendency for abstract conclusions to be stronger than the results section supports. In several fields this has been measured, and the mismatch is not rare.\n\nSo if a claim matters, check it against the results, not the abstract. This is the single highest-yield habit in reading research, and it takes one minute.\n\n## Note-taking that survives\n\nFor each paper, four lines in your own words:\n\n> **Claim.** What they say they showed.\n> **Evidence.** Design, n, key number, with the page or figure.\n> **Limitation.** What it cannot support.\n> **Use.** Which part of my argument this is for.\n\nWritten in your own words, because a copied summary is a copying-into-the-essay accident waiting to happen and because paraphrasing is where you find out whether you understood. Zotero holds these notes attached to the reference.\n\nThirty papers read this way is a working knowledge of a literature. Thirty AI summaries is a folder."},"check":{"question":"Why does the reading order put figures and captions second, before the introduction and methods?","options":[{"text":"Because figures are the part least likely to be affected by the authors' interpretation of their own data.","correct":false,"tempting":"Attributes objectivity to figures; how data are plotted, scaled and binned involves plenty of interpretation."},{"text":"Because methods sections are too technical to be useful until you already know what the result was, so they must come after everything else.","correct":false,"tempting":"Half right about sequencing but wrong about the reason for the figures' position, and it wrongly implies methods are always read last."},{"text":"Because the introduction is written last by most authors and therefore reflects the finished argument rather than the study.","correct":false,"tempting":"Offers a publishing-process claim that is often true and irrelevant here; the introduction's ordering is about triage cost, not sincerity."},{"text":"Because captions are usually written to stand alone, so the actual result is available in about ninety seconds.","correct":true}]},"prev":{"slug":"finding-the-paper-free","title":"Finding the paper, for nothing","url":"https://addaly.com/learn/ai-for-students/finding-the-paper-free"},"next":{"slug":"what-counts-as-evidence","title":"What counts as evidence","url":"https://addaly.com/learn/ai-for-students/what-counts-as-evidence"},"authorship":"Founder and series editor: Dr Mustafa Mun. Written by AI models to a written standard. Exam keys checked blind by a second model: 3,665 of 3,666 agreed. Academic review invited.","review":{"status":"Draft, open for academic review"},"cite":"Addaly (2026) 'Reading a paper without reading all of it', in AI for Students. Edited by M. Mun. Addaly. Available at: https://addaly.com/learn/ai-for-students/reading-a-paper-fast (Accessed: 13 October 2026).","licence":{"name":"CC BY-NC-SA 4.0","url":"https://creativecommons.org/licenses/by-nc-sa/4.0/"},"reader":{"schema":1,"title":"Reading a paper without reading all of it","courseTitle":"AI for Students","module":{"index":5,"slug":"research-and-evidence","title":"Research, sources and evidence"},"answer":null,"keyIdea":"Read in the order abstract, figures, aim, discussion, limitations, and check any claim that matters against the results section rather than the abstract, because abstracts systematically overstate.","hero":{"kind":"figure","figure":{"t":"figure","label":"Figure","figure":{"type":"flow","steps":[{"label":"Abstract","detail":"What they did and what they found. Decide whether to go on."},{"label":"Figures and captions","detail":"Usually the actual result. Good captions are written to be read alone."},{"label":"The aim","detail":"Last paragraph of the introduction, where the hypothesis lives."},{"label":"Discussion and limits","detail":"First and last paragraphs, then the limitations, where honest authors say what it cannot support."},{"label":"Results, to check","detail":"Does the number match what the abstract claimed? Sometimes it does not."}],"title":"The order researchers actually read in","caption":"Methods come last, and only if the result matters to you. The step almost nobody does is the final one: abstract conclusions have a documented tendency to be stronger than the results section supports, so a claim carrying weight in your essay gets checked against the results, which takes a minute. Stopping after the abstract is a successful outcome, not a failure."},"alt":"Figure: The order researchers actually read in. Steps: 1. Abstract (What they did and what they found. Decide whether to go on.); 2. Figures and captions (Usually the actual result. Good captions are written to be read alone.); 3. The aim (Last paragraph of the introduction, where the hypothesis lives.); 4. Discussion and limits (First and last paragraphs, then the limitations, where honest authors say what it cannot support.); 5. Results, to check (Does the number match what the abstract claimed? Sometimes it does not.). Methods come last, and only if the result matters to you. The step almost nobody does is the final one: abstract conclusions have a documented tendency to be stronger than the results section supports, so a claim carrying weight in your essay gets checked against the results, which takes a minute. Stopping after the abstract is a successful outcome, not a failure."}},"blocks":[{"t":"h2","id":"nobody-reads-papers-front-to-back","runs":[{"text":"Nobody reads papers front to back"}]},{"t":"p","runs":[{"text":"Researchers do not start at the first word and finish at the last. They triage. For an undergraduate reading twelve papers for an essay, the difference between triaging and not is the difference between finishing and not."}]},{"t":"p","runs":[{"text":"The order that works:"}]},{"t":"ol","start":1,"items":[[{"text":"Title and abstract.","b":true},{"text":" What did they do, what did they find. Decide whether to continue. Most papers stop here, and that is a successful outcome, not a failure."}],[{"text":"Figures and tables, with their captions.","b":true},{"text":" This is the surprising one. The figures usually contain the actual result, and captions in a good paper are written to be readable alone. You can often learn what a paper found in ninety seconds this way."}],[{"text":"The last paragraph of the introduction.","b":true},{"text":" Where the specific aim and hypothesis live."}],[{"text":"The first and last paragraphs of the discussion.","b":true},{"text":" What they think it means, and what they concede."}],[{"text":"The limitations section.","b":true},{"text":" Read this properly. It is where the honest authors tell you what the paper cannot support, and it is exactly what you need in order to cite it accurately."}],[{"text":"Methods","b":true},{"text":", only if the result matters to you. Then read it slowly, because this is where a study is strong or weak."}],[{"text":"Results text","b":true},{"text":", to check the numbers match what the abstract claimed. They sometimes do not."}]]},{"t":"h2","id":"where-the-model-helps-and-where-it-must-not","runs":[{"text":"Where the model helps and where it must not"}]},{"t":"p","runs":[{"text":"The safe pattern: read the paper yourself in that order, and use the model on the parts you cannot follow, grounded on the pasted text."}]},{"t":"quote","runs":[{"text":"Here is the methods section. Explain what a mixed-effects model is doing here and why they would have chosen it over a repeated-measures ANOVA. Use only what is in the text; say what is missing rather than filling it in."}]},{"t":"p","runs":[{"text":"That is a good use. It is a comprehension aid on material in front of you."}]},{"t":"p","runs":[{"text":"The unsafe pattern is asking for a summary and citing the summary. Two things go wrong. Summaries drop conditions and hedges (the "},{"text":"in older adults only","i":true},{"text":", the "},{"text":"when baseline was controlled","i":true},{"text":") and those qualifications are frequently the entire finding. And a summary you did not check gives you no page number and no sentence, so you cannot cite honestly."}]},{"t":"h2","id":"the-questions-to-ask-of-any-study","runs":[{"text":"The questions to ask of any study"}]},{"t":"p","runs":[{"text":"Whatever the field:"}]},{"t":"ul","items":[[{"text":"How many participants or observations, and how were they selected?","b":true},{"text":" A study of 24 undergraduates supports different claims from one of 24,000 adults."}],[{"text":"Compared with what?","b":true},{"text":" No control group means no causal claim. A comparison with nothing is not a comparison."}],[{"text":"Was anything randomised?","b":true},{"text":" This is the difference between "},{"text":"associated with","i":true},{"text":" and "},{"text":"caused","i":true},{"text":"."}],[{"text":"What exactly was measured?","b":true},{"text":" "},{"text":"Wellbeing","i":true},{"text":" measured how? Self-report on one item is not the same as a validated scale."}],[{"text":"How many analyses were run?","b":true},{"text":" Twenty tests produce a significant result by chance."}],[{"text":"Who funded it, and who benefits from the finding?","b":true},{"text":" Not disqualifying; relevant."}],[{"text":"Is this the first report or a replication?","b":true},{"text":" First reports are much more often wrong than the confidence of their abstracts suggests."}]]},{"t":"p","runs":[{"text":"A model can be asked to apply this list to a paper you have pasted, and that is a good use: it is applying a checklist to supplied text rather than recalling anything. Verify the answers against the paper."}]},{"t":"h2","id":"abstract-inflation","runs":[{"text":"Abstract inflation"}]},{"t":"p","runs":[{"text":"Worth knowing as a specific hazard: abstracts systematically overstate. They are written to be read and cited, they omit the qualifications, and there is a documented tendency for abstract conclusions to be stronger than the results section supports. In several fields this has been measured, and the mismatch is not rare."}]},{"t":"p","runs":[{"text":"So if a claim matters, check it against the results, not the abstract. This is the single highest-yield habit in reading research, and it takes one minute."}]},{"t":"h2","id":"note-taking-that-survives","runs":[{"text":"Note-taking that survives"}]},{"t":"p","runs":[{"text":"For each paper, four lines in your own words:"}]},{"t":"quote","runs":[{"text":"Claim.","b":true},{"text":" What they say they showed."}]},{"t":"quote","runs":[{"text":"Evidence.","b":true},{"text":" Design, n, key number, with the page or figure."}]},{"t":"quote","runs":[{"text":"Limitation.","b":true},{"text":" What it cannot support."}]},{"t":"quote","runs":[{"text":"Use.","b":true},{"text":" Which part of my argument this is for."}]},{"t":"p","runs":[{"text":"Written in your own words, because a copied summary is a copying-into-the-essay accident waiting to happen and because paraphrasing is where you find out whether you understood. Zotero holds these notes attached to the reference."}]},{"t":"p","runs":[{"text":"Thirty papers read this way is a working knowledge of a literature. Thirty AI summaries is a folder."}]}],"check":{"question":[{"text":"Why does the reading order put figures and captions second, before the introduction and methods?"}],"options":[{"text":[{"text":"Because figures are the part least likely to be affected by the authors’ interpretation of their own data."}],"correct":false,"tempting":[{"text":"Attributes objectivity to figures; how data are plotted, scaled and binned involves plenty of interpretation."}]},{"text":[{"text":"Because methods sections are too technical to be useful until you already know what the result was, so they must come after everything else."}],"correct":false,"tempting":[{"text":"Half right about sequencing but wrong about the reason for the figures’ position, and it wrongly implies methods are always read last."}]},{"text":[{"text":"Because the introduction is written last by most authors and therefore reflects the finished argument rather than the study."}],"correct":false,"tempting":[{"text":"Offers a publishing-process claim that is often true and irrelevant here; the introduction’s ordering is about triage cost, not sincerity."}]},{"text":[{"text":"Because captions are usually written to stand alone, so the actual result is available in about ninety seconds."}],"correct":true,"tempting":null}],"order":[2,3,1,0],"why":null},"prev":{"slug":"finding-the-paper-free","title":"Finding the paper, for nothing","minutes":9},"next":{"slug":"what-counts-as-evidence","title":"What counts as evidence","minutes":8}}}