2.3 Extracting Information from Data

What you need to know

• Information is the collection of facts and patterns extracted from data. Data becomes information when it's processed, organized, and interpreted.
• Programs make it possible to find patterns in datasets far too large for a person to read — this is why computing transformed fields from medicine to sports.
• Metadata is data about data: a photo's timestamp and location, a file's author, an email's routing headers. Metadata can be used to organize, search, and filter data, and changing or deleting it does not change the underlying data.
• Common operations for extracting information: filtering (keeping records that meet a condition), sorting, searching, aggregating (sum, average, count), and visualizing (charts, graphs) to make patterns visible.
• Combining data from multiple sources can reveal relationships not visible in any single source.
• Correlation is not causation. Two things trending together doesn't mean one causes the other. The exam tests this directly.
• Data can be incomplete, invalid, or biased — from how it was collected, who was included, or how it was cleaned. Conclusions inherit those flaws.
• The scale of data matters: large datasets may need distributed or parallel processing, and challenges include storage size, transfer time, and cleaning.
• Privacy concerns arise because metadata and combined datasets can identify or track people even when no single record seems sensitive.

Worked example

A school's attendance system stores 40,000 rows: student ID, date, period, present/absent. That's data. A program filters to first period, groups by day of week, and computes the absence rate — revealing Mondays have 30% more absences. That's information. If someone then claims "Monday causes absences," that's the correlation/causation trap; a sports schedule or bus route could be the real driver. The row's timestamp and the ID of the device that recorded it are metadata.

Going deeper

The nuance, edge cases, and connections that turn a 3 into a 5.

• The CED separates data (raw facts) from information (what you learn from analyzing data) and knowledge (what you can do with that information). A list of temperatures is data. "Average temperature rose 2° over ten years" is information. "We should plan for more cooling demand" is knowledge.
• Metadata has a precise role in the CED: it helps find, organize, and manage data. A photo library sorted by date uses metadata; searching email by sender uses metadata. Changing metadata (renaming a file) doesn't change the data (the file's contents).
• Metadata is also a privacy vector. A photo's GPS metadata reveals where you were. An email's header metadata reveals who you talk to and when, even if the message is encrypted. This is why "harmless" metadata can matter.
• Scalability of data processing: as datasets grow to millions or billions of records, single computers can't hold or process them in reasonable time. This is where parallel and distributed computing (Big Idea 4) enters — the two Big Ideas connect here.
• Incomplete or invalid data can come from collection errors, sensor failures, people skipping survey questions, or data corruption. The exam's point: conclusions drawn from flawed data are flawed, no matter how good the analysis.
• Data can be biased at collection (who was surveyed), at cleaning (which records were removed), and at visualization (a chart with a truncated axis exaggerates differences). Each stage is a place bias can enter.
• Combining datasets can reveal patterns, but it can also de-anonymize people: individually harmless datasets, when joined, can identify individuals. This is a real privacy concern the CED expects you to recognize.

Mistakes that cost points

• Accepting causation from correlation. The most common wrong answer in this topic. If two things trend together, the correct options say "correlated" or "a third factor may explain it." Options saying "X causes Y" are wrong unless the question describes a controlled experiment.
• Thinking metadata is part of the data. The date a document was created isn't in the document's text. The exam treats them as separate.
• Assuming more data fixes bias. If the collection method is biased, more data from the same method is more biased data. The fix is a better method, not more volume.
• Reading a chart without checking the axes. Truncated or uneven axes are a named form of misleading visualization. If a question shows a chart, look at the scale before drawing conclusions.

Practice questions

Written in the style of the real exam. Try each one before revealing the answer.