Module C.2 - Ethics Issues Identification
ModulesModule C.2 - Ethics Issues Identification
Module C introduces the AI Ethics Issues Identification process as the practical evaluation framework of the AIOLIA training programme. It translates the seven ALTAI requirements into an actionable, checklist-based methodology that enables learners to systematically detect, evaluate, and document ethical risks across real-world AI applications.
Building upon the core operational workflow established in Module B, this module anchors abstract ethics principles to concrete evaluation steps. Through guided theory, illustrative vignettes, and applied use case scenarios, it trains learners to move beyond passive awareness and rigorously analyze how complex issues—such as automation bias, proxy discrimination, data privacy erosion, black-box opacity, and loss of human agency—manifest across diverse operational contexts.
Learning Goals
- Master the 7 Assessment List for Trustworthy Artificial Intelligence (ALTAI) requirements and their core ethical dimensions.
- Apply Section 0 of the AIOLIA Issues Identification Checklist to define system scope, inputs/outputs, and affected stakeholder groups.
- Systematically evaluate AI systems across Sections 1–7 of the checklist using structured Likert-scale assessments and narrative scenario criteria.
- Distinguish between technical, organizational, and societal root causes of ethical risks within real-world deployments.
- Formulate clear, auditable issue statements to document identified risks, creating a reliable foundation for downstream mitigation strategies.
Ethics Issues Identification
The practical identification of AI ethics issues requires applying the seven Assessment List for Trustworthy Artificial Intelligence (ALTAI) requirements to concrete operational settings using the AIOLIA Issues Identification Checklist. The evaluation process begins by Establishing Operational Context (Section 0), where practitioners define system parameters, including input/output specifications, user profiles, decision autonomy, and affected stakeholder groups, before assessing risk. Grounding risk identification in specific operational contexts prevents vague, generic evaluations and ensures that downstream risk assessments reflect actual deployment realities.
Across Sections 1–7 of the checklist, practitioners systematically evaluate real-world scenarios to detect failure modes and ethical vulnerabilities. Evaluated risks include subtle loss of student agency caused by over-reliance on adaptive AI tutoring prompts, severe safety hazards stemming from data poisoning in industrial monitoring tools, and privacy violations arising from unverified training data in organizational assistants. The evaluation further examines black-box opacity in automated risk scoring platforms, algorithmic bias and proxy discrimination in content moderation systems, skills degradation among domain experts, and unclear legal liability during clinical decision-support failures.
To complete the identification process, qualitative observations and Likert-scale evaluations are synthesized into Structured, Auditable Issue Statements. Each statement explicitly documents the affected ALTAI requirement, the precise operational context, the root cause of the vulnerability, and the potential severity of harm. This standardized documentation creates a clear audit trail, establishing the necessary empirical foundation for selecting targeted technical and organizational mitigation measures.
Human Agency and Oversight
The Human Agency and Oversight requirement emphasises the importance of preserving meaningful human control over AI systems, particularly when their outputs influence decisions with significant consequences.
Although the ALTAI requirements provide a general framework for identifying ethics issues, they are non-exhaustive and should be interpreted in relation to the specific application context. The level and type of human oversight required may vary depending on the AI system, its intended purpose and the potential consequences of its outputs.
The following presentation explores this requirement through an AIOLIA use case involving safety engineers who use AI tools to identify hazards and assess risks. It illustrates how appropriate human oversight enables engineers to review, question and override AI-generated recommendations, while retaining responsibility for safety-related decisions.
A practical scenario involving an AI system that automatically modifies a safety analysis demonstrates what can go wrong when human intervention is not possible. It highlights the risks of losing control over a safety-critical process and allowing AI errors to propagate into safety requirements.
Finally, the presentation introduces questions from the AI Ethics Issues Identification Checklist to help identify potential issues related to human supervision, intervention and control.
As you explore the examples, consider whether humans have sufficient information, authority and practical means to oversee AI-supported processes and intervene when necessary.Technical Robustness and Safety
The Technical Robustness and Safety requirement emphasises the importance of ensuring that AI systems operate reliably, securely and safely throughout their lifecycle. AI systems should be resilient to errors, inconsistencies, vulnerabilities and attempts at misuse, while incorporating appropriate safeguards to prevent or minimise potential harm.
The following presentation explores this requirement through an AIOLIA use case involving safety engineers who use AI tools to support System-Theoretic Process Analysis (STPA) and accelerate software release approvals. It examines the integration of AI decision-support tools into safety analysis workflows, where reliability and accuracy are essential.
A practical scenario illustrates what can go wrong when an AI system processes corrupted data and produces inaccurate or inconsistent assessments without triggering appropriate fallback mechanisms. It highlights the risks associated with data poisoning, unreliable outputs and the absence of safeguards to prevent errors from propagating into safety-critical processes.
Finally, the presentation introduces questions from the AI Ethics Issues Identification Checklist to help identify potential issues related to cybersecurity, system misuse, reliability and the monitoring of false positives and false negatives.
As you explore the example, consider how the reliability and security of an AI system can affect the safety of the wider system in which it operates, and what safeguards are needed to detect, prevent and respond to failures.Privacy and Data Governance
The Privacy and Data Governance requirement emphasises the importance of protecting personal data and ensuring that data is collected, processed, stored and shared responsibly. Organisations should establish appropriate safeguards throughout the data lifecycle, ensuring that personal and sensitive information is handled lawfully and that individuals' rights are respected.
The following presentation explores this requirement through an AIOLIA use case involving users interacting with AI personal assistants that handle sensitive data. These systems process information about users' habits, routines, behavioural patterns and personal goals to provide personalised recommendations. This raises important questions about how personal data is collected, used and protected.
A practical scenario illustrates what can go wrong when sensitive personal data is stored without adequate protection and shared with third-party services without clear user consent or awareness. It highlights the risks of unauthorised data access, misuse of personal information and loss of user trust.
Finally, the presentation introduces questions from the AI Ethics Issues Identification Checklist to help identify potential issues related to data collection, dataset characteristics, privacy safeguards, consent and individuals' rights over their personal data.
As you explore the example, consider whether users are adequately informed about how their data is processed, whether appropriate safeguards are in place and whether they retain meaningful control over their personal information.Transparency
The Transparency requirement emphasises the importance of making AI systems understandable and their outputs accessible to relevant stakeholders. This includes providing appropriate information about the system's purpose, capabilities, limitations and decision-making processes, as well as ensuring that outputs can be traced, explained and meaningfully reviewed.
The following presentation explores this requirement through an AIOLIA use case involving employees assessed by AI-based cybersecurity systems. These systems analyse behavioural patterns, contextual risk factors and past interactions to assess employees' vulnerability to phishing attacks and recommend preventive actions.
A practical scenario illustrates what can go wrong when an employee receives a high-risk score without an adequate explanation of how it was generated. The absence of traceability and meaningful information about the data and factors influencing the assessment prevents the employee and relevant organisational stakeholders from understanding, verifying or challenging the result.
Finally, the presentation introduces questions from the AI Ethics Issues Identification Checklist to help identify potential issues related to the explainability of AI outputs, the documentation of system processes and the verification and validation of results.
As you explore the example, consider whether relevant stakeholders receive sufficient information to understand how AI-generated outputs are produced, assess their reliability and challenge decisions when necessary.Diversity, Non-discrimination and Fairness
The Diversity, Non-discrimination and Fairness requirement emphasises the importance of ensuring that AI systems treat individuals and groups fairly and do not produce discriminatory or unjust outcomes. This involves considering the diversity and representativeness of the data used, identifying potential biases and assessing how AI-generated outputs may affect different groups in specific contexts.
The following presentation explores this requirement through an AIOLIA use case involving security professionals using AI systems for content moderation. These systems help identify potentially harmful or illegal content, while human operators rely on their outputs to assess context, intent and potential harm.
A practical scenario illustrates what can go wrong when an AI moderation system reflects biases in its training data and disproportionately flags content associated with certain groups or cultural contexts. It highlights how differences in language, tone and origin can lead to inconsistent moderation decisions, the misclassification of legitimate expressions and unequal treatment of users.
Finally, the presentation introduces questions from the AI Ethics Issues Identification Checklist to help identify potential issues related to dataset representativeness, discrimination against specific groups and the fairness of AI-generated outputs.
As you explore the example, consider whether an AI system performs consistently across different groups and contexts, whose perspectives and experiences are represented in its data, and how biased or unfair outcomes can be identified and addressed.Societal and Environmental Well-being
The Societal and Environmental Well-being requirement emphasises the importance of considering the broader effects of AI systems on individuals, society and the environment. This includes assessing how AI systems may influence human behaviour, affect individual well-being and autonomy, and change the way people develop and exercise their skills and judgement.
The following presentation explores this requirement through an AIOLIA use case involving users interacting with AI personal assistants for daily life support. These systems provide personalised guidance on daily routines, habits and personal decisions through conversational interactions. While they may support users in organising their everyday lives, their influence on human behaviour and decision-making also needs to be considered.
A practical scenario illustrates what can go wrong when users become increasingly dependent on an AI assistant for everyday decisions. Over time, excessive reliance on its recommendations may weaken independent judgement and decision-making skills, particularly when the system does not encourage reflection or provide safeguards against over-reliance.
Finally, the presentation introduces questions from the AI Ethics Issues Identification Checklist to help identify potential issues related to users' skills and competencies, over-reliance on AI outputs, the preservation of human expertise and wider societal impacts.
As you explore the example, consider how AI systems can support individual well-being without undermining human autonomy, independent judgement and the skills needed to make informed decisions.Accountability
The Accountability requirement emphasises the importance of ensuring that appropriate responsibility is assigned for the development, deployment and use of AI systems. Organisations should establish clear roles and responsibilities, maintain records that support traceability and auditability, and provide mechanisms to investigate incidents, address harm and implement corrective actions.
The following presentation explores this requirement through an AIOLIA use case involving therapists using AI-generated deepfakes in psychotherapy. These systems simulate interactions with deceased loved ones or individuals associated with traumatic experiences, supporting therapeutic activities through realistic conversational experiences. Their use in such sensitive contexts raises important questions about responsibility, oversight and the handling of potential harm.
A practical scenario illustrates what can go wrong when an AI-generated interaction causes unexpected emotional distress and worsens a patient's condition. It highlights the difficulties that arise when responsibilities are not clearly assigned among therapists, system developers and the organisation deploying the tool, and when mechanisms for tracing system decisions and responding to incidents are inadequate.
Finally, the presentation introduces questions from the AI Ethics Issues Identification Checklist to help identify potential issues related to the assignment of responsibility, the traceability and auditability of AI systems, and the involvement of ethics expertise in AI development and deployment.
As you explore the example, consider who should be responsible when an AI system contributes to harm, how responsibilities should be defined across the different stakeholders, and what mechanisms are needed to investigate incidents and ensure appropriate responses.Short scenarios
Read each scenario carefully. Select all the ethics requirements that are relevant to the issues described. More than one answer may be correct.
