IPA IP 2025 Question 8

Topic: Business Industry

A question IPA Japan published in 令和7年度 ITパスポート試験 公開問題 問8, translated into English. ITPEC is based on the same exams.

Among the descriptions concerning the handling of data used in machine learning of AI, which of the following lists all and only the appropriate measures that can be considered for reducing bias and ensuring the quality of data?

AConfirm that the data is suitable for the purpose of training.BConfirm the source and creation history of the data.CPerform the annotation of data in line with the purpose of training.DExclude data for image recognition that is considered difficult to identify even with the human eye.

Explanation

All four measures — each one keeps unsuitable, untraceable, mislabelled or ambiguous data out of training.

A model learns whatever its training data contains, including skews and errors, so data is checked before it is used:

  • •A: data that does not match the purpose (other conditions, a skewed population) teaches the model the wrong thing or a biased view.
  • •B: knowing where the data came from and how it was made (its provenance) shows whether it can be trusted, whether it is biased, and whether it may be used.
  • •C: annotation attaches the correct answers (labels) the model learns from; labels set by criteria that fit the purpose and are applied consistently keep the data accurate.
  • •D: an image that even a person cannot identify cannot be labelled reliably, so it adds noise rather than knowledge.

Answer (b)

Why not others:

  • •(a) — leaves out C and D, the checks on the labels and on ambiguous images
  • •(c) — leaves out B and C; provenance and consistent annotation matter as much
  • •(d) — leaves out A; data unsuited to the purpose is a main source of bias

Key rule: Training data quality decides model quality: check that the data fits the purpose, know its provenance, label it consistently for the purpose, and drop items that cannot be labelled reliably.

Business Industry · AI-written, may contain errors

Practice this exam

This question comes from the past questions IPA Japan publishes, translated into English by ITPEC Practice. ITPEC Practice is an independent study tool and is not affiliated with IPA or ITPEC. See the original paper (PDF) or Report an issue.