Short Answer

Why Human Approval Does Not Prove Synthetic Data Works for a Model

Question: Why is it not enough for a person to say that generated transaction records look realistic in order to conclude that they are realistic for a machine learning model?

Sample answer: A person may judge realism by surface appearance, while a model depends on statistical details such as distributions, relationships between variables, and rare-value patterns. Because synthetic data can seem plausible to a person yet still fail to match the patterns a model needs, human approval alone is not sufficient.

Key points:

  • People and models evaluate realism using different criteria.
  • Data can look plausible to a person but still be unsuitable for a model.

Rubric: The response must explain that human judgment and model judgment differ, and that data can appear realistic to a person without having the statistical properties needed for machine learning.

0

1

Updated 2026-08-12

Contributors are:

Who are from:

Tags

Machine Learning

Deep Learning

Supervised Learning

Dive into Deep Learning @ D2L

Data Science

Machine Learning Strategy

Machine Learning Yearning @ DeepLearning.AI