Part 2 | Putting it to work
Chapter 11. Using AI for data analysis
Key points
- Visualising survey, spreadsheet and assessment data becomes easy
- It lowers the doorway into analysis, which used to be high
- But the fine detail of statistics needs a specialist's eye
You can load a file and analyse it
Generative AI also works as a support for data analysis. Upload
- survey results
- Google Forms responses
- Excel data
- CSV files
- assessment data
- research data
and it will help with means, aggregation, charts and spotting tendencies. It suits "I just want to see the shape of it" particularly well.
Start by making it visible
Have it lay out
- age distribution
- satisfaction
- pain scores
- activity levels
- response proportions
as bar charts, pie charts, scatter plots and tables. Then adjust in words:
- "make it easier to read"
- "change the colours"
- "simpler"
What used to mean wrestling with Excel operations, formulas and chart settings now happens in natural language.
It fits therapy work well
Therapists handle more data than they realise:
- patient satisfaction surveys
- study-group feedback forms
- before-and-after comparisons
- cognitive function data
- activity data
- usage logs from digital devices
And it is easy to end up never doing anything with it, because the aggregation is tedious. AI is very useful as a way to at least get a rough view.
Statistics call for care
Statistical processing does need caution. AI can produce
- wrong interpretations
- an inappropriate choice of method
- confident bluffing
Be especially careful with
- interpreting p-values
- multivariate analysis
- regression
- claims of causation
For genuine research, conference presentations or journal submissions, always involve a statistician or supervisor and check the original sources.
Start with simple analysis
- means and standard deviations
- response proportions
- cross-tabulation
- a rough sense of correlation
Treating these as "for reference" is enough at the start. For example:
- "Does exercise frequency look related to satisfaction?"
- "Does sleep duration look related to fatigue?"
- "Which age group uses this most?"
Use it as a doorway into seeing tendencies. It adapts beyond research to improving how you work, to teaching, and to running a community.
Personal information, again
The same caution applies. Names, addresses, phone numbers, IDs, hospital and facility names, detailed histories — remove or anonymise them before entering anything. Keeping a running sense of what may and may not go into an AI matters.
Something that lowers the doorway
Data analysis used to carry a strong impression of being difficult. With generative AI, at least
- charting something
- looking at the tendency
- asking a question about it
is now within anyone's reach. In other words, AI is
something that lowers the doorway into analysis
A prompt to try
You are a researcher who is good at data analysis. Read the attached (anonymised) survey data and analyse it in this order.
- Lay out respondent characteristics (age, sex, profession) as a table and charts.
- Give simple aggregation (proportions) for each question.
- Cross-tabulate "satisfaction" against the other items and pick out those that look related.
- List three cautions or limits on interpretation.
At the end, comment on whether a proper statistical analysis is needed.
Chapter 11 summary
- Making survey and assessment data visible becomes far easier
- The fine detail of statistics must be reviewed by a specialist
- Anonymise personal information without exception