Bot-filtered vaping posts are not a verified group of people
Trace account screening and retained text separately from confirmed human identity, personal use and observed quitting outcomes.
No Vape · Updated

Locate the account decision in the data flow
Write the collection rule first, then the screening steps and final material analyzed. Distinguish an account-level classification from a rule removing an unrelated post or a duplicate text. One account can contribute many posts; an account retained by a screening rule does not become several independently observed people because it posted frequently.
An original 2017 JMIR Public Health and Surveillance study used BotOrNot account scores when examining e-cigarette Twitter data and compared corpora including and excluding accounts identified as social bots. It also described manual removal of unrelated spam. We cite those methods only; their topic findings and historical counts are not current estimates or evidence about an account you encounter today.
Keep a model label separate from verified identity
Find which information the model used and how the authors chose their inclusion rule. A score is an output of that procedure, not a personal identity document. A retained account might still have an uncertain operator or purpose. Conversely, a flagged account is not proof that a named person deliberately manipulated a conversation.
If the paper reports evaluation of screening performance, retain its reference and conditions. Do not convert a general accuracy claim into certainty about every retained account or infer the error count in this dataset without evidence. This reading exercise does not select thresholds, investigate private profiles or recommend running a detector on fellow community members.
Try a collection-and-identity note
Imagine a fictional headline saying that real quitters endorsed an idea after a bot filter was applied to public posts. Your note asks three separate questions: which accounts were retained by the rule, whether their operators were verified, and whether any quitting outcome was observed. An answer to the first cannot silently supply the other two.
Keep the source’s actual noun, such as retained posts or accounts classified under the stated method, instead of upgrading it to confirmed patients or successful quitters. The example has no real accounts, numerical results or allegation. Its purpose is to locate exactly where the headline adds an unsupported identity or outcome.
Preserve what the filtered analysis can still contribute
A filtered dataset can support a bounded account of the analyzed discussions. It need not be dismissed because identity is not fully known. Equally, it does not establish population agreement or validate a medical claim in a post. Return to the text-selection and analysis methods for the claim you actually need to understand.
This differs from screening repeated submissions to a survey: here the issue is the inferred kind of account producing public material. It also differs from a topic model, which groups text patterns. End with the account rule, its effect on included material and the identity or behavior question left unresolved. For your own support, use a genuine service or an agreed personal contact rather than treating a research screening label as someone’s credentials.
Sources and scope
Sources inform the guidance. The worksheets and examples are No Vape editorial suggestions, not tested treatments.
Written with AI assistance and checked against the cited sources. No clinical review is claimed. This guide covers practical planning, not individual medical treatment. How we research and write