Branchoria Think Before Sharing Think Before Sharing

Practical judgement for social feeds, AI answers, and viral claims.

Within Think Before Sharing

When Research Becomes Cherry Picking

Responsible research means using a method for weighing evidence, not collecting links that confirm a preferred conclusion.

On this page

  • Effort versus method
  • Confirmation traps
  • A better personal research routine
Preview for When Research Becomes Cherry Picking

Introduction

“Do your own research” is good advice only when it means using a fair method for weighing evidence. It becomes cherry-picking when someone starts with the answer they prefer, searches until they find links that support it, and treats the search itself as proof of diligence. In the age of social media and AI, this distinction matters because online tools can make one-sided evidence feel abundant, authoritative and personally discovered.

Overview image for Own Research Responsible personal research is not about collecting the most screenshots, threads, videos or search results. It is about asking a claim to survive contact with better sources, opposing evidence, original context and uncertainty. Research on online misinformation shows why this is harder than it sounds: people can be pulled towards familiar or belief-confirming material, search results can reflect the wording of a biased query, and even attempts to verify false news online can sometimes increase belief in it when the surrounding search results are low quality. [Nature+2First Monday]nature.comOnline searches to evaluate misinformation can increase…by K Aslett · 2024 · Cited by 170 — We present consistent evidence that…

Effort Is Not the Same as Method

A person can spend hours “researching” and still come away less informed if the process is mostly confirmation. The internet rewards visible effort: long threads, many tabs, saved clips, quoted posts and confident summaries. But effort only helps if it is attached to a method that tests the claim rather than protects it.

The phrase “do your own research” often sounds empowering because it rejects blind trust. That is useful when it pushes people to check a source, read beyond a headline or compare claims across independent outlets. The problem begins when the phrase becomes a licence to treat every institution as suspect except the few sources that already agree with the researcher’s preferred conclusion. At that point, “research” becomes a story of personal awakening rather than a discipline of evidence.

This is not just a problem of individual stubbornness. Search itself can be shaped by the way a question is phrased. An audit of Google Scholar and Semantic Scholar found that confirmation-biased academic queries could return results aligned with the bias embedded in the query, with disparities varying by topic and platform. That matters because even academic-looking search results are not neutral piles of truth; they are ranked outputs responding to the words the user supplied. [First Monday]firstmonday.orgFirst MondayAn algorithm audit of Google and Semantic Scholarby C Kacperski · 2024 · Cited by 12 — This study examines whether confirmati…

A better test is to ask: “What would I have to find for this claim to become weaker?” If the answer is “nothing”, the activity is not research. It is defence.

How Confirmation Traps Work Online

Confirmation bias is the tendency to search for, interpret, prefer or remember information that supports existing beliefs. Online, that bias does not need to operate as a conscious decision. It can appear in small choices: which search terms feel natural, which headline looks worth clicking, which expert seems “captured”, which correction feels suspicious, and which anecdote feels more real than a dataset.

Social platforms intensify the trap because content usually arrives with social cues: likes, reposts, comments, reaction videos and group approval. A claim can begin to feel credible because many people appear to be engaging with it, even before the reader has checked whether it is true. Research on misinformation sharing has also found that people often share more accurately when their attention is directed towards accuracy, suggesting that social media environments can distract users from truth-checking rather than simply reveal a lack of reasoning ability. [PMC]pmc.ncbi.nlm.nih.govPMCAccuracy prompts are a replicable and generalizableby G Pennycook · 2022 · Cited by 363 — Online misinformation has become a major focus of attention in recent years among academics, te…

Search engines add a second trap: the user can mistake finding something for confirming something. A search for “evidence that X is a scam” is likely to produce a different information path from “independent reviews of X”, “regulatory action against X”, or “best evidence for and against X”. The first query may still return real pages, but those pages are filtered through a question that already leans towards one conclusion.

The most dangerous version is the “answer-first” search:

  • Start with a conclusion: “This is fake”, “This cure is being hidden”, “This policy caused everything”.
  • Use loaded search terms: “cover-up”, “truth about”, “exposed”, “debunked lie”, “what they are not telling you”.
  • Select matching results: choose the pages, clips or posts that support the starting belief.
  • Dismiss contradictory evidence: treat disagreement as proof of corruption, bias or censorship.
  • Declare independence: present the chosen material as “my own research”.

The flaw is not that the person searched. The flaw is that the search was designed to confirm.

Own Research illustration 1

Why “Just Google It” Can Backfire

A striking finding from recent misinformation research is that searching online to evaluate false news can sometimes make people more likely to believe it. A Nature study across five experiments found consistent evidence that online search to assess the truthfulness of false news articles increased belief in those false articles, especially when users encountered lower-quality search results. [Nature]nature.comOnline searches to evaluate misinformation can increase…by K Aslett · 2024 · Cited by 170 — We present consistent evidence that…

This does not mean verification is pointless. It means verification has to be done carefully. Searching a suspicious claim immediately after it appears can lead into a weak information environment where copies, partisan commentary, low-quality blogs or manipulated pages appear before strong reporting or expert correction has caught up.

Data & Society’s work on “data voids” explains one reason this happens. A data void occurs when a search query has limited, non-existent or poor-quality results, leaving space for manipulators to fill the gap with misleading material. Obscure phrases, breaking events, newly coined slogans and oddly specific claims can all create conditions where the first available results look more authoritative than they deserve. [Data & Society]datasociety.netdata voids where missing data can easily be exploitedData & SocietyData Voids: Where Missing Data Can Easily Be Exploited11 May 2018 — Data Voids: Where Missing Data Can Easily Be Exploited…Published: May 2018

This is especially relevant in social media and AI environments. A viral post may include a distinctive phrase; thousands of people search that phrase; the available results may be dominated by people repeating or reacting to the same claim. The reader then sees many pages apparently “confirming” the claim, when in reality they may be seeing duplication, not independent corroboration.

The practical lesson is simple: do not verify a claim using only the vocabulary supplied by the claim-maker. Change the wording. Search for the underlying event, named institutions, original document, location, date, image source or exact quote. A claim that survives multiple neutral search paths is stronger than one that appears only when searched in its own loaded language.

AI Makes Cherry-Picking Look Cleaner

Generative AI can help organise research, but it can also make cherry-picking more polished. A chatbot can summarise one-sided sources into balanced-sounding prose, produce confident explanations without visible uncertainty, and generate references that look credible until checked. UNESCO’s 2025 media and information literacy work emphasised that AI can make mistakes and that critical judgement remains necessary when engaging with AI-created content. [UNESCO]unesco.orgAI can make mistakes: Why media literacy matters more…24 Oct 2025 — media and information literacy (MIL) can help people think c…

AI also changes the emotional feel of research. Instead of scrolling through messy pages, a user receives a fluent answer in seconds. That fluency can create a false sense of closure: the topic feels settled because the answer reads smoothly. But a well-written answer is not the same as a well-supported answer.

There is growing evidence that hallucinated or non-existent citations are a real problem in AI-assisted knowledge work. A 2026 large-scale preprint auditing 111 million references across major repositories estimated a sharp rise in non-existent references after widespread large language model adoption, including a conservative estimate of 146,932 hallucinated citations in 2025 alone. [arXiv]arxiv.orgOpen source on arxiv.org.

For personal research, the implication is not “never use AI”. It is: never let AI be the final authority on whether its own claims are true. Use AI for brainstorming search terms, identifying possible counterarguments, summarising documents you can inspect, or explaining unfamiliar concepts. For factual claims, original sources still matter.

Hallucination-Free? Assessing the Reliability of Leading AI Legal Research Tools (Paper Explained)

Channel: Yannic Kilcher · Views: 42.1K · Uploaded: June 2024 · Length: 1 hour 11 minutes

Open on YouTube

A Better Personal Research Routine

Good research does not require professional training, but it does require a routine that makes cherry-picking harder. The aim is to move from “Can I find support for this?” to “What is the best available evidence, and how much confidence does it justify?”

One widely taught approach is lateral reading: leaving the page you are on and checking what other credible sources say about the source, author, organisation or claim. Stanford’s Civic Online Reasoning materials describe lateral reading as investigating who is behind an unfamiliar online source by opening new tabs and seeing what trusted sources say about it. [Inquiry Group]cor.inquirygroup.orgOpen source on inquirygroup.org.

Mike Caulfield’s SIFT method gives a compact version of this habit: stop, investigate the source, find better coverage, and trace claims, quotes or media back to their original context. University library guides and media literacy programmes continue to teach SIFT because it turns research into a sequence of checks rather than a vibe-based judgement. [University of Chicago Library Guides]guides.lib.uchicago.eduUniversity of Chicago Library Guides The SIFT MethodUniversity of Chicago Library Guides The SIFT Method

A practical routine can look like this:

  1. State the claim plainly.

Reduce the post, video or AI answer to one checkable sentence. “This politician is corrupt” is too broad. “This politician received £50,000 from Company X in March 2025” is checkable.

  1. Separate the claim from the reaction.

Anger, fear, humour and identity signals are not evidence. Note them, but do not let them decide the conclusion.

  1. Use neutral search terms first.

Search names, dates, documents, locations and exact phrases. Avoid starting with loaded words such as “hoax”, “cover-up” or “truth”, unless you are deliberately checking how a narrative is being promoted.

  1. Read laterally before deeply.

Before spending twenty minutes on a single page, check who runs it, what its track record is, whether other credible sources refer to it, and whether it has a reason to distort the topic.

  1. Find the strongest opposing evidence.

Do not settle for the weakest critic of your view. Look for the best argument or evidence against the conclusion you currently favour.

  1. Trace important claims backwards.

If a post cites a report, find the report. If an article quotes a study, open the study or at least its abstract and methods summary. If a video uses a clip, look for the longer original.

  1. Classify confidence, not just truth.

Some claims are false. Some are true but exaggerated. Some are plausible but unproven. Some are impossible to assess from public evidence. A good researcher is comfortable saying “unclear”.

This routine is slower than scrolling, but often faster than being misled and having to rebuild a belief later.

Own Research illustration 2

What Counts as Stronger Evidence?

Not all sources do the same job. A personal testimony may be useful for understanding someone’s experience, but weak for estimating how common that experience is. A peer-reviewed study may be strong for a narrow question, but not automatically decisive for policy. A government page may be authoritative on official rules, but not neutral on political interpretation. Good research weighs sources by what they are capable of proving.

For online claims, the most useful evidence often has at least one of these qualities:

  • Primary documentation: original reports, court filings, datasets, legislation, company statements, official statistics, full interview transcripts or archived pages.
  • Independent corroboration: multiple credible sources reaching similar findings without simply copying one another.
  • Transparent methods: clear explanation of how evidence was gathered, what was excluded, and what uncertainty remains.
  • Relevant expertise: people or institutions with demonstrated knowledge of the specific topic, not just general confidence or popularity.
  • Correction culture: sources that publish corrections, link to evidence, distinguish reporting from opinion, and show their work.

The reverse is also important. Weak evidence often relies on screenshots without provenance, anonymous claims without context, edited clips, AI-generated summaries with no traceable sources, or chains of articles that all point back to the same unverified origin.

A useful personal rule is to ask: “What would this source know, and how would it know it?” A neighbour may know what happened on their street. They may not know the national cause of a trend. A scientist may understand a study’s methods. They may not know whether a viral clip from another country is authentic. Evidence becomes stronger when the source’s access matches the claim being made.

Lateral Reading

Channel: MCPS-MD · Views: 1.8K · Uploaded: September 2024 · Length: 3 minutes 18 seconds

Open on YouTube

Avoiding False Balance Without Cherry-Picking

Avoiding cherry-picking does not mean pretending all sides are equally credible. Sometimes the evidence is lopsided. Climate change, vaccine safety, election administration, public health and conflict reporting all include claims where a fringe narrative may be loud but poorly supported. Fair research does not require giving weak claims equal weight for the sake of appearing open-minded.

The difference between fair weighting and cherry-picking is method. Fair weighting asks which explanation best fits the total evidence. Cherry-picking asks which evidence can be made to fit the preferred explanation.

This distinction matters because “both sides” can become its own trap. A reader may collect one mainstream source and one fringe source, then assume the truth lies halfway between them. But evidence does not work like a seesaw. A claim supported by many independent lines of evidence should not be treated as equal to a claim supported by speculation, misread statistics or a single viral anecdote.

At the same time, majority agreement is not a substitute for checking. The right response is proportional confidence: stronger evidence earns more confidence; weaker evidence earns less; uncertainty remains visible. UNESCO frames media and information literacy as a way to engage critically with information, navigate digital environments safely and build trust in information ecosystems, not as a habit of reflexive distrust. [UNESCO]unesco.orgOpen source on unesco.org.

How to ‘Do Your Own Research’ with Medical Journalist Dr. Trisha Pasricha (Part 2)

Channel: LinkedIn News · Views: 251 · Uploaded: March 2025 · Length: 35 minutes

Open on YouTube

The Social Cost of Bad Research

Cherry-picked research rarely stays private. It becomes a post, a family argument, a group chat warning, a workplace rumour, a political identity marker or a reason not to trust a legitimate institution. The harm is not only that one person holds a false belief; it is that poor research habits scale through networks.

Social media gives cherry-picked evidence an advantage because it is easy to package. A screenshot of one paragraph travels better than a careful explanation of uncertainty. A confident thread feels more satisfying than a mixed conclusion. A short AI-generated summary can remove caveats that were present in the original sources. By the time a correction appears, the first claim may already have shaped what people searched for, remembered and shared.

This is why responsible research is a civic skill, not just a private virtue. In a high-speed information environment, the question is not “Have I found a source?” but “Have I checked this well enough to believe it, act on it or pass it on?”

Own Research illustration 3

The Practical Standard: Try to Disprove Yourself

The best everyday defence against cherry-picking is not to become perfectly neutral. No one is. The better goal is to build a habit of self-challenge: before accepting a claim, make one serious attempt to find out why it might be wrong, overstated or missing context.

That standard changes the feel of research. It turns searching from a hunt for ammunition into a stress test. It encourages neutral search terms, lateral reading, original sources and fair comparison. It also makes AI more useful because the user can ask for counterarguments, missing evidence and source checks rather than simply requesting a persuasive answer.

A claim that survives this process may still be uncertain, but it has earned more trust than a claim supported only by convenient links. In the age of social media and AI, “doing your own research” is responsible only when it includes the discipline to look for what would change your mind.

Amazon book picks

Further Reading

Books and field guides related to When Research Becomes Cherry Picking. Use these as the next step if you want deeper reading beyond the article.

Browse more on Amazon: Thinking, Fast and Slow Calling Bullshit Factfulness

As an Amazon Associate I earn from qualifying purchases.

eBay marketplace picks

Marketplace Samples

Example marketplace items related to this page. Use the search link to explore similar finds on eBay.

Using USA
Browse more on eBay.co.uk

Example items shown for inspiration; availability and pricing can change. Branchoria may earn a commission if you purchase through outbound eBay links.

Endnotes

  1. Source: nature.com
    Link: https://www.nature.com/articles/s41586-023-06883-y
    Source snippet

    Online searches to evaluate misinformation can increase...by K Aslett · 2024 · Cited by 170 — We present consistent evidence that...

  2. Source: arxiv.org
    Link: https://arxiv.org/abs/2403.17286

  3. Source: pmc.ncbi.nlm.nih.gov
    Title: PMCAccuracy prompts are a replicable and generalizable
    Link: https://pmc.ncbi.nlm.nih.gov/articles/PMC9051116/
    Source snippet

    by G Pennycook · 2022 · Cited by 363 — Online misinformation has become a major focus of attention in recent years among academics, te...

  4. Source: unesco.org
    Link: https://www.unesco.org/en/articles/ai-can-make-mistakes-why-media-literacy-matters-more-ever
    Source snippet

    AI can make mistakes: Why media literacy matters more...24 Oct 2025 — media and information literacy (MIL) can help people think c...

  5. Source: arxiv.org
    Link: https://arxiv.org/abs/2605.07723

  6. Source: unesco.org
    Link: https://www.unesco.org/en/media-information-literacy

  7. Source: unesco.org
    Title: strengthens media and information literacy across ukraine
    Link: https://www.unesco.org/en/articles/unesco-strengthens-media-and-information-literacy-across-ukraine

  8. Source: unesdoc.unesco.org
    Link: https://unesdoc.unesco.org/ark%3A/48223/pf0000388547

  9. Source: unesco.org
    Link: https://www.unesco.org/mil4teachers/sites/default/files/medias/fichiers/2023/10/Part_1_CI%20MIL_CURRICULUM.pdf

  10. Source: unesco.org
    Link: https://www.unesco.org/en/articles/unesco-launches-free-online-course-tackle-climate-disinformation-through-media-and-information

  11. Source: unesdoc.unesco.org
    Link: https://unesdoc.unesco.org/ark%3A/48223/pf0000392035

  12. Source: unesco.org
    Link: https://www.unesco.org/en/articles/ai-meets-governance-bhutan-civil-servants-trained-ai-and-media-information-literacy

  13. Source: unesco.org
    Title: feature conference
    Link: https://www.unesco.org/en/weeks/media-information-literacy/feature-conference

  14. Source: unesdoc.unesco.org
    Link: https://unesdoc.unesco.org/ark%3A/48223/pf0000395373

  15. Source: arxiv.org
    Link: https://arxiv.org/html/2403.17286v1

  16. Source: arxiv.org
    Link: https://arxiv.org/html/2504.13777v1

  17. Source: nature.com
    Link: https://www.nature.com/articles/s41467-022-30073-5

  18. Source: nature.com
    Link: https://www.nature.com/articles/s41562-024-02086-1

  19. Source: stacks.stanford.edu
    Title: Civic Online Reasoning National Portrait
    Link: https://stacks.stanford.edu/file/gf151tb4868/Civic%20Online%20Reasoning%20National%20Portrait.pdf

  20. Source: ed.stanford.edu
    Title: it doesn t take long learn how spot misinformation online stanford study finds
    Link: https://ed.stanford.edu/news/it-doesn-t-take-long-learn-how-spot-misinformation-online-stanford-study-finds

  21. Source: scholar.google.com
    Link: https://scholar.google.com/citations?hl=en&user=AIbJenwAAAAJ

  22. Source: scholar.google.com
    Link: https://scholar.google.com/citations?hl=en&user=NJwhbHoAAAAJ

  23. Source: firstmonday.org
    Link: https://firstmonday.org/ojs/index.php/fm/article/view/13730/11709
    Source snippet

    First MondayAn algorithm audit of Google and Semantic Scholarby C Kacperski · 2024 · Cited by 12 — This study examines whether confirmati...

  24. Source: datasociety.net
    Title: data voids where missing data can easily be exploited
    Link: https://datasociety.net/library/data-voids-where-missing-data-can-easily-be-exploited/
    Source snippet

    Data & SocietyData Voids: Where Missing Data Can Easily Be Exploited11 May 2018 — Data Voids: Where Missing Data Can Easily Be Exploited...

    Published: May 2018

  25. Source: cor.inquirygroup.org
    Link: https://cor.inquirygroup.org/curriculum/lessons/intro-to-lateral-reading/

  26. Source: guides.lib.uchicago.edu
    Title: University of Chicago Library Guides The SIFT Method
    Link: https://guides.lib.uchicago.edu/c.php?g=1241077&p=9082322

  27. Source: Wikipedia
    Title: Confirmation bias
    Link: https://en.wikipedia.org/wiki/Confirmation_bias

  28. Source: datasociety.net
    Title: Data Society Data Voids Final 3
    Link: https://datasociety.net/wp-content/uploads/2018/05/Data_Society_Data_Voids_Final_3.pdf

  29. Source: cor.inquirygroup.org
    Title: teaching lateral reading
    Link: https://cor.inquirygroup.org/curriculum/collections/teaching-lateral-reading/

  30. Source: cor.inquirygroup.org
    Title: lateral vs vertical reading
    Link: https://cor.inquirygroup.org/curriculum/lessons/lateral-vs-vertical-reading/

  31. Source: cor.inquirygroup.org
    Link: https://cor.inquirygroup.org/

  32. Source: pmc.ncbi.nlm.nih.gov
    Link: https://pmc.ncbi.nlm.nih.gov/articles/PMC9038198/

  33. Source: pmc.ncbi.nlm.nih.gov
    Link: https://pmc.ncbi.nlm.nih.gov/articles/PMC9082967/

  34. Source: pmc.ncbi.nlm.nih.gov
    Link: https://pmc.ncbi.nlm.nih.gov/articles/PMC11393549/

  35. Source: pmc.ncbi.nlm.nih.gov
    Link: https://pmc.ncbi.nlm.nih.gov/articles/PMC9761900/

  36. Source: pubmed.ncbi.nlm.nih.gov
    Link: https://pubmed.ncbi.nlm.nih.gov/38123685/

  37. Source: pmc.ncbi.nlm.nih.gov
    Link: https://pmc.ncbi.nlm.nih.gov/articles/PMC12964124/

  38. Source: pmc.ncbi.nlm.nih.gov
    Link: https://pmc.ncbi.nlm.nih.gov/articles/PMC12425422/

  39. Source: uscupstate.libguides.com
    Title: Lateral Reading
    Link: https://uscupstate.libguides.com/LateralReading

  40. Source: mediahelpingmedia.org
    Title: lateral reading
    Link: https://mediahelpingmedia.org/basics/lateral-reading/

  41. Source: libguides.ucmerced.edu
    Title: sift method
    Link: https://libguides.ucmerced.edu/news/evaluation/sift-method

  42. Source: GOV.UK
    Title: main report cdei review into bias in algorithmic decision making
    Link: https://www.gov.uk/government/publications/cdei-publishes-review-into-bias-in-algorithmic-decision-making/main-report-cdei-review-into-bias-in-algorithmic-decision-making

  43. Source: ojs.aaai.org
    Link: https://ojs.aaai.org/index.php/AAAI/article/view/42257/46218

  44. Source: scribbr.com
    Title: confirmation bias
    Link: https://www.scribbr.com/research-bias/confirmation-bias/

  45. Source: scienceofboosting.org
    Title: Lateral Reading
    Link: https://www.scienceofboosting.org/project/lateral-reading/

Additional References

  1. Source: youtube.com
    Link: https://www.youtube.com/watch?v=PYyNqPLfu2o
    Source snippet

    Non-Scientific Knowledge and Doing Your Own Research (Confirmation Bias and the Internet Age)...

  2. Source: youtube.com
    Title: Overthinking About Doing Your Own Research
    Link: https://www.youtube.com/watch?v=VTNR6maU5Ag
    Source snippet

    How to 'Do Your Own Research' with Medical Journalist Dr. Trisha Pasricha (Part 2)...

  3. Source: researchgate.net
    Link: https://www.researchgate.net/publication/330710636_Fighting_misinformation_on_social_media_using_crowdsourced_judgments_of_news_source_quality

  4. Source: researchgate.net
    Link: https://www.researchgate.net/publication/349535570_Lateral_reading_College_students_learn_to_critically_evaluate_internet_sources_in_an_online_course

  5. Source: researchgate.net
    Link: https://www.researchgate.net/publication/360253696_Accuracy_prompts_are_a_replicable_and_generalizable_approach_for_reducing_the_spread_of_misinformation

  6. Source: coachingforleaders.com
    Link: https://coachingforleaders.com/podcast/prevent-being-duped-mike-caulfield/

  7. Source: prosocialdesign.org
    Link: https://www.prosocialdesign.org/library/[accuracy-prompts

  8. Source: brookings.edu
    Link: https://www.brookings.edu/articles/algorithmic-bias-detection-and-mitigation-best-practices-and-policies-to-reduce-consumer-harms/

  9. Source: ntu.ac.uk
    Link: https://www.ntu.ac.uk/studenthub/news/2025/10/navigating-ai-and-misinformation

  10. Source: newslit.org
    Link: https://newslit.org/checkology-resources/

Topic Tree

Follow this branch

Parent topic

Think Before Sharing

Related pages 24

More on this topic 6