03 · What You Need to Know
Understanding Social Media Content as Documentary Evidence
Why Can Social Media Posts Qualify as Documents?
Social media posts are recorded artifacts that can preserve statements, representations, interactions, and other forms of communication.
For example, a university's official announcement about permitted generative AI use may provide direct evidence of how the institution publicly communicates its expectations.
A public discussion among teachers may provide evidence of how particular platform users articulate concerns about AI-assisted assessment.
These materials can be examined through documentary analysis because their recorded content is itself relevant to the phenomenon.
However, their documentary status does not determine whether they can be collected or reproduced without restrictions.
The distinction between what qualifies as a research document and what constitutes ethically permissible research use is essential.
What Types of Social Media Materials May Be Analyzed?
Depending on the research question, relevant materials may include public posts, comments, institutional announcements, discussion threads, images, captions, hashtags, or recorded interactions.
For example, a researcher studying university AI governance communication may analyze official posts announcing policy changes.
Another researcher may examine public discussions about how teachers interpret those changes.
However, the analytical unit must be defined clearly.
A post, comment, thread, account, and institutional campaign are not interchangeable units.
Researchers should specify which materials are included and whether contextual information from surrounding interactions is necessary for interpretation.
Public Accessibility and Ethical Publicness Are Different
A post may be technically accessible without the author reasonably anticipating its use in a published research study.
For example, a person may publish a comment in a publicly viewable discussion while primarily addressing a small community of peers.
Extracting that comment, combining it with other information, and reproducing it in an academic publication may change its audience and potential consequences.
The Association of Internet Researchers' ethical guidance emphasizes contextual judgment rather than treating the public-private distinction as a simple binary.
Researchers should consider the communication setting, audience expectations, identifiability, and possible harms.
Technical accessibility alone should not determine the ethical decision.
Are Official Institutional Accounts Different from Personal Accounts?
Official institutional accounts and personal accounts may have different documentary functions and privacy implications.
An official university account generally communicates on behalf of an organization to an intended public audience.
Its posts may therefore be particularly relevant to research about institutional communication or public policy representation.
Personal accounts may contain statements about experiences, frustrations, professional relationships, or sensitive circumstances.
Although these statements may be publicly accessible, their use can create greater risks of identification or unwanted exposure.
Researchers should assess the origin, purpose, and intended audience of each source category rather than applying identical assumptions to every account.
Does Public Social Media Research Always Require Individual Consent?
Not universally.
Consent requirements depend on the nature of the research, the data collected, identifiability, reasonable privacy expectations, applicable law, and institutional ethics requirements.
Some studies of public institutional communications may not involve human participants in the same way as studies examining identifiable personal experiences.
Other projects may require consent, additional safeguards, or formal review even when the material is publicly visible.
Researchers should not assume that a public account removes all obligations or that every social media study necessarily requires contacting every author.
The appropriate determination should follow the relevant ethical and regulatory framework.
What Does Contextual Integrity Mean for Social Media Research?
Contextual integrity concerns whether information flows remain appropriate to the social context in which information was originally shared.
Nissenbaum (2004) argues that privacy cannot be understood solely through secrecy or unrestricted access.
For example, a teacher may publicly discuss classroom difficulties with a professional community while not expecting their comments to be permanently archived and scrutinized in an identifiable academic publication.
Researchers should consider how collecting, aggregating, quoting, or republishing material changes its circulation.
This does not mean that all public online research is impermissible.
It means that ethical evaluation should consider the social expectations surrounding the original communication.
Can Public Posts Be Quoted Verbatim?
Verbatim quotation can create identification risks because distinctive phrases may be searchable.
Removing usernames does not necessarily prevent readers from locating the original post.
For example, a researcher might reproduce a teacher's unusual description of an AI-related classroom incident without naming the account.
A search for the exact wording could nevertheless reveal the author's identity.
Researchers should evaluate whether exact quotation is necessary for the analytical purpose and whether the potential identification risk is acceptable.
Paraphrasing, aggregation, or obtaining appropriate permission may be considered depending on the context.
However, paraphrasing should preserve the substantive meaning and should not be presented as a verbatim quotation.
What About Screenshots?
Screenshots can preserve visual layout, timestamps, interaction context, and other information that may be lost when text is extracted.
They may be useful when the research concerns visual communication or platform presentation.
However, screenshots can also expose usernames, profile images, account identifiers, and unrelated participants.
Researchers should consider whether the visual details are necessary and whether identifying information can be appropriately protected.
Copyright and platform conditions may also affect reproduction.
A screenshot should not automatically be included in a publication merely because it was publicly accessible when collected.
How Do Platform Features Shape Documentary Evidence?
Social media content is produced within platforms that structure visibility, interaction, and communication.
Character limits, recommendation systems, reply structures, moderation, and account settings may influence what users publish and what researchers can retrieve.
For example, short posts may simplify complex opinions, while discussion threads may distribute meaning across several contributions.
Researchers should avoid treating individual posts as self-contained statements when their meaning depends on surrounding exchanges.
Platform-specific features should be documented when they materially affect the analysis.
Findings should not be generalized across platforms without appropriate evidence.
Does Social Media Content Represent Public Opinion?
Not necessarily.
People who post publicly may differ systematically from those who remain silent, use private accounts, or do not participate on the platform.
Highly visible posts may also reflect platform ranking and engagement mechanisms rather than the distribution of views in a broader population.
For example, frequent criticism of generative AI in a selected online discussion does not establish that most teachers oppose its use.
Researchers should distinguish the opinions expressed in the sampled content from the prevalence of opinions among teachers, students, or the general public.
Population-level claims require an appropriate sampling frame and inferential design.
How Should a Social Media Document Corpus Be Defined?
The corpus should be bounded by explicit criteria concerning platform, account type, topic, language, period, visibility, and unit of analysis.
For example, a study might include official public posts from selected university accounts announcing generative AI policies during a specified academic year.
Another study might sample public discussion threads containing defined terms within a specified period.
These designs support different conclusions.
The procedures for defining a document corpus should account for platform-specific retrieval limitations and changing content.
Researchers should explain how eligible material was identified and what remained inaccessible.
What Sampling Problems Are Specific to Social Media?
Social media sampling may be affected by incomplete search results, deleted posts, restricted accounts, platform ranking, language differences, and changes to data-access systems.
For example, collecting only posts returned by a platform's search interface may not produce a complete record of all eligible discussions.
Researchers should distinguish the target population of posts from the content actually retrievable through the selected method.
Hashtag sampling may also exclude relevant posts that discuss the same issue without using the selected hashtag.
These limitations should inform the scope of inference.
Convenient access to a large volume of posts does not establish representative coverage.
How Should Deleted or Edited Posts Be Handled?
Social media content may be edited, deleted, hidden, or made private after initial publication.
Researchers should establish procedures for recording retrieval dates and relevant version information.
However, retaining deleted material may raise ethical questions, especially when an author has attempted to withdraw public access.
The appropriate response depends on the research purpose, applicable rules, and potential risks.
Researchers should not assume that historical availability creates an unrestricted right to preserve and republish personal content indefinitely.
Where possible, the treatment of changed or deleted material should be considered before collection begins.
Can Automated Collection Be Used?
Automated collection may be technically feasible for some platforms, but its permissibility and completeness depend on current platform conditions, available interfaces, and applicable legal requirements.
Researchers should verify authorized access procedures rather than assume that public visibility permits unrestricted scraping.
Automated collection can also introduce methodological problems if the retrieved data omit replies, context, deleted content, or relevant metadata.
Collection scripts should be documented sufficiently to explain what was retrieved and under which conditions.
Technical efficiency does not establish ethical acceptability or evidentiary completeness.
How Do You Distinguish Human and Automated Accounts?
Social media datasets may contain institutional accounts, automated accounts, coordinated campaigns, and individual users.
These sources may produce similar-looking posts while representing different communication processes.
For example, repeated statements about an educational policy may result from automated reposting rather than independent expressions of opinion.
Researchers should determine whether account type matters to the research question.
However, identifying automated behavior can be uncertain, and automated-account detection methods may produce classification errors.
Unverified assumptions about account authenticity should not be presented as established findings.
How Should Social Media Content Be Analyzed?
Social media documents may be analyzed using qualitative content analysis, thematic analysis, discourse analysis, quantitative content analysis, or other appropriate approaches.
For example, a qualitative study might examine how teachers frame responsibility for verifying AI-generated materials.
A quantitative study might classify the frequency of specified themes across eligible institutional posts.
The choice between quantitative, qualitative, and mixed-methods content analysis should follow the research question and characteristics of the corpus.
Researchers should also decide whether posts, comments, threads, accounts, or institutions constitute the relevant analytical units.
Can Likes, Shares, and Comments Be Treated as Measures of Public Support?
Engagement indicators may provide evidence about recorded platform interactions, but they should not automatically be interpreted as endorsement or agreement.
A post may receive attention because users disagree with it, find it controversial, or share it critically.
Platform algorithms and account visibility may also influence engagement.
Researchers should specify what an engagement metric measures operationally and avoid treating it as a direct measure of attitudes without supporting evidence.
The distinction between observable platform activity and inferred psychological states is essential.
What About Sensitive Topics and Vulnerable Participants?
Additional safeguards may be necessary when social media content concerns sensitive experiences, minors, stigmatized conditions, or potentially harmful disclosures.
For example, publicly posted student complaints about academic misconduct investigations may contain identifiable information with consequences for the individuals involved.
Researchers should consider whether collecting or quoting such material could increase exposure or harm.
Public visibility should not be interpreted as informed consent to all subsequent research uses.
Ethical decisions should reflect contextual expectations, foreseeable consequences, and applicable institutional requirements.
How Should Social Media Data Be Stored?
Data-management procedures should reflect the sensitivity, identifiability, and permitted uses of the collected material.
Researchers may need to separate identifiers from analytical records, restrict access, establish retention periods, and protect downloaded files.
However, anonymization may be difficult when exact quotations or contextual details remain searchable.
Researchers should distinguish removing direct identifiers from preventing reidentification.
Platform conditions may also restrict redistribution of collected content.
The data-management plan should address these issues before collection and publication.
When Are Additional Data Sources Necessary?
Social media posts can establish what selected accounts publicly communicated, but they may not establish why users posted, whether statements reflect enduring beliefs, or how widely those views are held.
For example, a teacher's criticism of AI policy on social media does not necessarily provide a complete account of their professional experience.
Interviews may help examine personal interpretations, while surveys may support population-level estimates when appropriately designed.
However, additional methods are not automatically required when the research question concerns the public discourse itself.
The need for triangulation with other data sources depends on the claims the study intends to make.
How Should Social Media Documentary Research Be Reported?
The methodology should identify the platform, relevant period, account or content eligibility criteria, retrieval procedures, and analytical units.
Researchers should describe the achieved corpus, known access limitations, and treatment of edited or deleted content.
Ethical reporting should explain relevant consent or review arrangements, protection of identities, and quotation practices.
Where platform conditions may affect representativeness, the implications should be acknowledged.
Readers should be able to understand what the data represent without being led to assume that public social media activity is equivalent to the opinions of a broader population.
