The concern here revolves aroundpotential algorithmic biasandpreferential treatment, which falls under the core ethical principle ofFairness and Non-Discrimination.
According to ISO/IEC 42001:2023, organizations mustidentify and mitigate discriminatory impactsof AI systems, especially where algorithmic decisions influenceuser experiences, accessto services, or market fairness(Clause 6.1.2 and Clause 8.2.3).
This core element ensures that AI decisions are:
* Equitable and unbiased
* Do not systematically favor or disadvantage individuals or groups
* Transparent in rationale and decision logic
The PECB Guide emphasizes that Fairness is particularly important inrecommendation, ranking, or classification systems, where outcomes affect stakeholders' access or exposure.

Machine Learning (ML)is the most suitable AI concept in this scenario. ML focuses on developing algorithms that canlearn from structured or unstructured dataand make predictions based on historical patterns.
In this case, analyzing customerbrowsing history and purchase recordsfalls directly undersupervised learning, a subcategory of ML, which is typically used forpredictive modelingin retail (such as next-best- offer, product recommendation, or demand forecasting).
According to the PECB Lead Auditor Study Guide (Domain 1),ML is specifically referenced as the core techniquefor prediction systems, user behavior modeling, and data-driven decision-making systems.
Though Deep Learning (DL) is a subset of ML, it is often used for more complex pattern recognition tasks such as image or speech recognition, which is not explicitly required here.

The most appropriate AI concept for analyzingimages and videosisComputer Vision. Computer Vision is a subfield of artificial intelligence that enables systems tointerpret and process visualdata, such as photos and video frames, which is exactly what is required in this scenario.
According to thePECB Lead Auditor Guide,Computer Visionis explicitly associated with tasks such as object recognition, content moderation, facial recognition, and image classification - all of which are relevant in detecting inappropriate content on platforms like social media.
WhileDeep Learningis often usedwithinComputer Vision (e.g., convolutional neural networks), thecorrect high-level conceptbeing asked here is Computer Vision, which encompasses the overall domain applicable to this scenario.
* NLPis used for analyzing text and language, not visual content.
* MLis a broader category under which Computer Vision models are trained, but is too general for this specific task.

The correct core element isAccountability.
According toISO/IEC 42001:2023 - Clause 5.3, andPECB Lead Auditor Guide - Domain 1,accountability is defined as the obligation of organizations totake responsibility for the outcomes and impactsof AI systems across the lifecycle - including design, development, deployment, and operation.
This includes establishing:
* Clear roles and responsibilities
* Oversight mechanisms
* Escalation procedures for unintended consequences
It directly addresses the ethical and governance need to ensure AI systems are not used irresponsibly or without clear attribution.

The correct term here isAI System Impact Assessment(AIIA), which is distinctly referenced inISO/IEC
42001:2023 - Clause 6.1.2as part of the organization's process to identify and assesspotential impactsof AI systems on stakeholders.
An AIIA is designed to evaluate theethical, societal, legal, and human rights implicationsof AI use. It supportstransparency, stakeholder trust, and ethical alignment.
WhileAI Risk Assessment(Clause 6.1.1) focuses more on organizational and system-level risks (e.g., technical, legal), theImpact Assessmentlooks atexternal consequences- especially forindividuals and groups.