Skip to main content

What is pattern recognition?

The AI Glossary

In our classes, we often pause to explain the ideas behind AI and the language used to describe it. These posts grew out of those conversations. They’re for anyone who would like to understand AI better, with no technical background assumed.

Pattern recognition means identifying useful regularities in information. A system might recognize a handwritten character, identify a recurring sound or distinguish different kinds of messages. The pattern is a relationship among observed features; recognizing it does not necessarily explain why it exists.

A simple example is reading handwritten digits. Different people write the same number differently. A useful recognition system needs to tolerate those variations while distinguishing, say, a three from an eight. It may learn from labeled examples or use rules designed by people.

Pattern recognition overlaps with machine learning but is not identical to it. Machine learning is a way of developing behavior from data. Pattern recognition describes a family of tasks. Some pattern-recognition systems are based on fixed rules, while some machine-learning tasks concern choosing actions or generating content rather than assigning a familiar category.

The representation matters. Software does not receive a photograph as a human experience; it receives numerical measurements. The selected measurements or learned representations determine which similarities are easy to detect. Two images can resemble one another mathematically for a reason that is irrelevant to the intended question.

This is where shortcuts arise. A system intended to recognize a finding may instead rely on a background marker that happened to accompany it in the training examples. The shortcut can work well on familiar data and fail when the background changes. A high score on one test collection does not reveal every distinction the system has learned.

In an outpatient practice, recurring combinations of delayed laboratory feeds, unacknowledged results and missed follow-up calls might help identify abnormal test results at risk of going unaddressed. Recognizing that pattern is useful only if the team understands what it represents. Are patients hard to reach, are routing rules misconfigured, or are acknowledgments being recorded inconsistently? The same visible pattern can have different causes.

Pattern recognition can direct attention and support prediction. Establishing cause, deciding what to do and checking whether an intervention helps require further work. Those are not reasons to dismiss the pattern; they are reasons to be precise about what has been discovered.

Sources: Hugging Face — Image Classification; Hugging Face — Text Classification

Further reading from Harvard Chan: Public Health Perspectives — To Meet Future Needs, Health Care Leaders Must Look at the Data (Science) (2022).

To learn more, take a look at the AI in Health Care Certificate of Specialization from the Advanced Learning Academy at Harvard T.H. Chan School of Public Health. You can take an individual course or work toward the full certificate.


I’m a doctor, not a statistician. I teach the AI in Health Care Certificate of Specialization with Dr. Heather Mattie, a statistician and co-director of Harvard Chan’s Health Data Science master’s program. For more on data science and statistics, see Harvard Chan’s Health Data Science resources.

AI is a fast-moving field. This explanation reflects our class discussions and related material at the time of writing. Terminology, capabilities and tools may change.

For suggested additions or corrections, please email me at tpanch@hsph.harvard.edu.

Dr. Trishan Panch
Program Director, AI in Health Care
Harvard T.H. Chan School of Public Health — Advanced Learning Academy

Lead with AI in Health Care

The AI Pragmatist is a monthly newsletter from ALA faculty, with new explainers & essays from the Advanced Learning Academy