5.6.3 Conflicting Requirements of Stakeholders in the Development of AI Systems

Introduction

The textbook explains that AI application developers have to consider and manage many conflicting requirements of stakeholders to avoid failure of the resulting software. In simple words, when building an AI system, different stakeholders want different things. These wants often clash with each other . Developers must find ways to balance these conflicting requirements.

Definition

Conflicting Requirements of Stakeholders refers to the situation where different stakeholders in an AI system have different, opposing, or incompatible needs, expectations, and goals that developers must manage.

Main Idea

The main idea is that conflicts among stakeholders are unavoidable in AI development. These conflicts arise because different stakeholders have: Different expectations Different desires Different goals Different values Different interpretations of the problem Developers must resolve these conflicts, not suppress them, to avoid failure of the AI system.

Core Concept

Why Conflicts Occur The textbook explains several reasons for conflicts: Diverse Expectations: Stakeholders of an AI application in a specific area (healthcare, education, supply chain, marketing) 1. may have many diverse expectations, desires, and goals in conflicting situations. Different Views: Stakeholders express their views or disagreements on some decisions or values proposed in an AI 2. application. Similar Needs, Different Implementation: Different stakeholders may have similar needs but may differ on how to 3. implement them. Different Interpretations: Conflicts can be due to different interpretation or understanding of the given problem.4. Key Principle The textbook states: "Conflicts among the stakeholders should be resolved, not suppressed." This means: Do not ignore conflicts Do not force one stakeholder's view on others Find balanced solutions Communicate openly Negotiate and compromise Why This Matters The textbook states: "Conflicting requirements of stakeholders make the developers' task of AI application design more difficult as they have to satisfy all the stakeholders." Challenge: Developers must satisfy all stakeholders. Risk: If conflicts are not managed, the AI system may fail.

Detailed Explanation

Types of Conflicts 1. Value Conflicts: Stakeholders have different ethical or moral values. Example: End users want privacy Developers want data for training Conflict: Privacy vs. Data access 2. Goal Conflicts: Stakeholders have different objectives. Example: Owners want profit Regulators want safety Conflict: Profit vs. Safety 3. Priority Conflicts: Stakeholders prioritize different things. Example: Customers want quick delivery Employees want manageable workload Conflict: Speed vs. Workload 4. Implementation Conflicts:

Stakeholders agree on the goal but disagree on how to achieve it. Example: All want fair AI Some want strict rules Others want flexible guidelines Conflict: Strict vs. Flexible 5. Interpretation Conflicts: Stakeholders understand the problem differently. Example: Developers see it as a technical problem Users see it as a social problem Conflict: Technical vs. Social Real-World Example AI in Healthcare Stakeholders and Their Requirements: Stakeholder Requirement Patients Accurate diagnosis, privacy, affordable care Doctors Decision support, not replacement, control

Hospital Efficiency, cost reduction, good outcomes AI Developers Access to data, time to develop, resources Regulators Safety, transparency, compliance Insurance Risk assessment, cost prediction Society Better healthcare, lower costs, fairness Conflicts: Patients vs. Developers: Patients want privacy; developers need data. Doctors vs. Hospital: Doctors want control; hospital wants automation. Hospital vs. Regulators: Hospital wants cost reduction; regulators want safety. Patients vs. Insurance: Patients want affordable care; insurance wants risk assessment. Developers vs. Regulators: Developers want flexibility; regulators want strict rules. How to Resolve Conflicts 1. Open Communication: Talk to all stakeholders Understand their perspectives Share information transparently 2. Stakeholder Engagement:

Involve stakeholders in design Conduct consultations Use participatory design methods 3. Negotiation and Compromise: Find middle ground Balance competing interests Seek win-win solutions 4. Ethical Frameworks: Use ethical principles to guide decisions Consider fairness, justice, and human rights Apply frameworks like utilitarianism or deontology 5. Transparent Decision-Making: Document decisions Explain rationale Be accountable 6. Prioritization: Identify critical requirements Balance must-haves vs. nice-to-haves Consider long-term impact 7. Iterative Development: Build, test, and refine Get feedback from stakeholders

Adjust based on input The Developer's Role Developers play a critical role in managing conflicts: Responsibilities: Identify Stakeholders: Know who is affected.1. Understand Interests: Learn what each stakeholder wants.2. Anticipate Conflicts: Predict where conflicts may arise.3. Facilitate Communication: Bring stakeholders together .4. Find Solutions: Develop balanced solutions.5. Implement Ethically: Ensure the system is fair and responsible.6. Monitor and Adjust: Continuously improve based on feedback.7.

Intermediate Level

In a business setting, stakeholder conflicts are common. Example: AI in Hiring Stakeholders and Their Requirements: Stakeholder Requirement Job Applicants Fair treatment, transparency, feedback Hiring Managers Efficient screening, quality candidates HR Department Compliance, diversity, legal protection AI Developers Access to data, time to build, clear requirements Regulators Non-discrimination, explainability Company Cost savings, better hires, competitive advantage

Conflicts: Applicants want transparency; developers want proprietary algorithms. HR wants diversity; AI might perpetuate bias. Company wants efficiency; regulators want explainability. Resolution: Use explainable AI Train on diverse data Regular audits for bias Transparent communication with applicants

Important Points

• Conflicting requirements are unavoidable in AI development.1.
• Conflicts arise from different expectations, desires, goals, and interpretations.2.
• Conflicts should be resolved, not suppressed.3.
• Developers must satisfy all stakeholders.4.
• Conflicts make development difficult but must be managed.5.
• Figure 5.11 shows stakeholders conflict in the development of AI system.6.
• Resolution strategies: Communication, engagement, negotiation, ethical frameworks.7.
• Developer's role: Identify, understand, anticipate, facilitate, solve, implement, monitor .8.

Remember

• Conflicting requirements = different stakeholders want different things
• Conflicts are unavoidable
• Resolve, don't suppress conflicts
• Reasons for conflicts: Different expectations, desires, goals, interpretations
• Stakeholders in AI: End users, developers, customers, suppliers, employees, regulators, funders, owners, policy makers,
• community
• Figure 5.11 = stakeholders conflict in AI development
• Resolution: Communication, engagement, negotiation, ethical frameworks
• Developer's role: Identify, understand, anticipate, facilitate, solve, implement, monitor
• Warning: Ignoring conflicts can lead to AI system failure

Common Mistakes

• Ignoring conflicts – Leads to AI system failure.1.
• Suppressing conflicts – Conflicts should be resolved.2.
• Favoring one stakeholder – All stakeholders should be considered.3.
• Not communicating – Open communication is essential.4.
• Assuming conflicts will resolve themselves – They need active management.5.
• Focusing only on technical aspects – Social and ethical aspects matter .6.
• Not involving stakeholders in design – Participatory design is important.7.

Quick Revision

Conflicting requirements of stakeholders in AI development refer to the situation where different stakeholders have different, opposing, or incompatible needs, expectations, and goals. Conflicts are unavoidable and arise from different expectations, desires, goals, and interpretations. The textbook states that conflicts should be resolved, not suppressed. Developers must satisfy all stakeholders, which makes their task difficult. Figure 5.11 shows stakeholders conflict in the development of AI system. Resolution strategies include communication, engagement, negotiation, and ethical frameworks. The developer's role includes identifying, understanding, anticipating, facilitating, solving, implementing, and monitoring stakeholder requirements.

60-Second Revision

Conflicting requirements = different stakeholders want different things Conflicts are unavoidable Resolve, don't suppress conflicts Reasons for conflicts: Different expectations, desires, goals, interpretations Stakeholders in AI: End users, developers, customers, suppliers, employees, regulators, funders, owners, policy makers, community Figure 5.11 = stakeholders conflict in AI development Resolution: Communication, engagement, negotiation, ethical frameworks Developer's role: Identify, understand, anticipate, facilitate, solve, implement, monitor Warning: Ignoring conflicts can lead to AI system failure

Example: AI in healthcare has many conflicting requirements