Five stars at two in the morning
A new gadget has a rating of 4.4 stars. Look closer. Five of the five-star reviews came from one account, in under two minutes, in the middle of the night.
Somebody made those on purpose. A good analyst can spot this, and keeps it out of the analysis.
Key words
| Word | What it means | Example |
|---|---|---|
| Evidence | facts, tests or counts that back up a claim | |
| Citation | a note that says where a fact or figure came from | a link to the survey report |
| Fake news | false stories or figures made to look real | |
| Malicious data | data made or changed on purpose to mislead or to harm | |
| Bot | a program that acts like a person online, often to post in bulk | |
| Duplicate | a record that appears more than once | |
| Filter | to take out the records that should not be counted, and to keep a note of why |
The cells students fill in are left blank in this preview.
By the end of this lesson
Note
I can check a source for evidence and citations, and spot and filter out fake or malicious data.
Starter
Have a go
Starter
3 questionsYour answers are marked here and are not kept. A college account keeps every student’s marks.
Facts need evidence and a citation
A good source shows its evidence: counts, tests or records. A citation says where each figure came from, so you can look it up.
A claim with no citation is only an opinion. Big claims with no evidence are a warning sign.
Strong and weak evidence
| Claim | Weak version | Strong version |
|---|---|---|
| Shop visits are rising | "Everyone knows shops are busier." | "Visits rose from 4,000 to 4,600 in a year (Table 3, shop counter log)." |
| A tool is fast | "It is the fastest ever." | "It ran the same job in 12 seconds in our test, 31 for the old tool." |
| Students like the menu | "Students love it." | "Of 120 students asked, 90 chose a hot meal (survey, 4 March)." |
The cells students fill in are left blank in this preview.
Fake news and malicious data
Fake news is a false story or figure that looks real. Malicious data is data made or changed on purpose to mislead or harm. Examples are fake reviews, a bot posting a thousand likes, a doctored screenshot, or a file with numbers quietly changed.
If you use it, your analysis is wrong, and so is every decision built on it.
Run it: who posted the reviews?
select account, count(*) as reviews, avg(stars) as avg_stars
from reviews
group by account
order by reviews desc;Warning signs, and what to do
| Warning sign | Check | Action |
|---|---|---|
| No source for a figure | Search for the original | Do not use until found |
| Many posts from one new account | Group the records by account | Filter them out and note why |
| Copied text in many records | Look for duplicates | Keep one, note the rest |
| A value that cannot be true | Compare with a known range | Mark it as bad data |
| A story only one site tells | Cross-check with other sources | Treat as unproven |
The cells students fill in are left blank in this preview.