The training of an ML model... What type of bias is LEAST important to look for when testing the model?

Question

Options

• AInappropriate bias
• BAutomation bias
• CAlgorithmic bias
• DSample bias

• A
  14% (5)
• B
  75% (27)
• C
  3% (1)
• D
  8% (3)

Explanation

The ISTQB CT-AI syllabus distinguishes between several types of bias relevant in AI testing, includingsample bias,algorithmic bias, andinappropriate bias. In Section3.3 - Bias in AI-Based Systems, the syllabus stresses the importance of identifying biases that originate fromtraining data,model development, anddecision logic. Sample bias occurs when the training data does not adequately represent the population; algorithmic bias arises when the model produces systematically skewed results due to learned patterns; inappropriate bias involves ethically or socially problematic distortions in the outcomes. All three of these bias types directly affect theoutputs of the AI modeland are therefore highly relevant when testing an industrial inspection system intended to reliably detect defects. These biases can lead to defective items being missed or false alarms being raised, which impacts quality assurance significantly . Automation bias, however, is fundamentally different. It refers to ahuman cognitive bias, where users (e.g., inspectors) overly trust or rely on the AI system's output. While important in user- interaction testing, it isnota biaswithin the ML model itself. Since the question asks which bias isleast important when testing the model, automation bias can be legitimately deprioritized duringmodel-level testing. Therefore, Option B is correct.

No community discussion yet for this question.