More Related Content
Similar to Presentation[1]---------------------.pdf (20)
Recently uploaded (20)
Presentation[1]---------------------.pdf
- 1. PROFESSIONAL ETHICS PRESENTATION Dr. / Dina Saif Ragab
- 2. INTERPRETABILITY TRANSPARENCY IN AI 1.Definition of Transparency Common Meaning: The ability to see through something clearly or explain a process openly. In AI: Combines two aspects understanding how decisions are made and justifying those decisions ethically. KEY ASPECTS OF AI TRANSPARENCY EXPLAINS HOW THE AI SYSTEM WORKS AND WHY IT BEHAVES A CERTAIN WAY. Known as "opening the black box" of AI. Ensures the system is clear and understandable. JUSTIFIES THE PROCESSES BEHIND THE AI'S DESIGN AND ITS OUTCOMES. Ensures decisions are ethical, fair, and trustworthy. JUSTIFIABILITY
- 3. WHY TRANSPARENCY IS IMPORTANT : Builds trust between AI systems and users. Ensures fairness and non-discrimination in AI decisions Provides accountability for errors or unethical outcomes 3 CRITICAL TASKS FOR DESIGNING AND IMPLEMENTING TRANSPARENT AI 1.Process Transparency, Task 1: Justify Process 2.Outcome Transparency, Task 2: Clarify Content and Explain Outcome 3.Outcome Transparency, Task 3: Justify Outcome
- 4. Process Transparency, Task 1: Justify Process. To ensure process transparency, stakeholders must see that ethical considerations, fairness, and public trust were prioritized throughout the AI system's design and implementation. This involves adhering to best practices across the AI lifecycle and establishing strong auditability measures using an accountability-by-design framework. These steps build confidence in the system's integrity and decision- making process.
- 5. Outcome Transparency, Task 2: Clarify Content and Explain Outcome. Outcome explanations should be simple and accessible to non-specialists, focusing on the reasoning behind a models decisions in a socially meaningful way. Instead of relying on technical or mathematical details, the explanation should connect to real-world practices and societal factors, ensuring stakeholders understand the decisions impact. Example: If an AI denies a loan application, rather than saying "credit score below 100," explain: "The loan was declined because the financial history shows repeated late payments, suggesting challenges in future repayments."
- 6. OUTCOME TRANSPARENCY, TASK 3 :JUSTIFY OUTCOME ETHICAL PERMISSIBILITY The systems decisions and actions align with ethical standards, ensuring no harm to individuals or society. NON- DISCRIMINATIO N/FAIRNESS The system is designed to avoid bias, ensuring equal treatment and opportunities for all individuals, regardless of race, gender, or other protected characteristics. SAFETY/PUBLIC TRUSTWORTHIN ESS The system incorporates safeguards to ensure user safety, data security, and reliability, fostering public trust in its integrity and accountability.
- 7. Purpose: This framework outlines key principles and actionable steps to ensure transparency, trust, and ethical justifiability in AI systems. Core Justifiability Criteria: Ethical Permissibility Promotes human dignity, autonomy, and flourishing. Aligns with values like justice, public good, and human connection. Non-Discrimination / Fairness Ensures fairness in data, design, outcomes, and implementation. Aims to prevent discriminatory harm. Safety / Public Trust Focuses on accuracy, reliability, security, and robustness. Supports sustainability in AI operations. Explicability Provides clear, understandable explanations. Yields socially meaningful and interpretable outcomes. Key Transparency Processes: Process Transparency: Stakeholder Impact Assessments and re-assessments. Fairness Statements and Dataset Factsheets. Testing, verification, and monitoring of AI safety. Outcome Transparency: Development of interpretable AI for clear communication. Justified outcomes through normative explanations.
- 8. (PBG) Framework: ETHICAL PERMISSIBILITY 1. Professional and Institutional Transparency Maintain rigorous standards of conduct and accountability for all team members.. NON- DISCRIMINATIO N/FAIRNESS 2. Clear and Accessible PBG Framework Define explicit policies and protocols governing the ethical use and design of AI. . SAFETY/PUBLIC TRUSTWORTHINESS 3. Robust Auditability and Logging Protocols Establish activity logging mechanisms to track key decisions, actions, and outcomes across the project lifecycle
- 9. Integrity ETHICAL GOVERNANCE 1. Professional Conduct Standards: Team members should adhere to rigorous conduct standards to ensure professionalism and institutional transparency throughout the AI project lifecycle. 2.Core Values: These standards should include values such as integrity, honesty, neutrality, objectivity, and impartiality, guiding all professionals involved. 1. Serving the Public Interest: Professionals in AI development must prioritize the public interest over any other concerns, in line with core civil service values. 2. Transparency and Public Scrutiny: The design and implementation process should be transparent and open to public scrutiny, with reasonable protection of sensitive information. Accountability
- 10. Governance and Auditability in AI Development. Purpose of a PBG Framework:. 1)Provides structured integration of values into AI workflows 2)Supports CRISP-DM, KDD, models. why use a BGB framework? 1)Overview of governance structures 2)Timeframes and follow-up actions 3)Highlights team roles and workflow stages 4)End-to-end auditability protocols Key Stages in the AI Workflow: Problem Formation Data Extraction & Acquisition Data Processing Modeling Testing & Validation Deployment Monitoring Reassessment
- 11. Enabling Auditability with a Process Log. 1)Ensure end-to-end auditability 2)Maintain records and track activities. 3)Collect data across all phases (modelling, training, testing, etc.). 4)Enhances transparency, supports compliance, improves workflows. Summary : _PBG Framework ensures responsible AI innovation. _Auditability provides transparency and accountability.
- 12. OUR TEAM Everest Cantu Ceo Of Ingoude Company Drew Holloway Ahmed Mohamed Hassan 4221320 Yousef Ahmed Abdelrahman 4221244 Remas Saad Mahmoud 4221275 Nour Ali 4221601 Salma Hazem Mostafa 4221280 Foad Reda Foad 4221041 Elsayed Mohamed Ali 4211268 Marwan Ahmed Foad 4241810 Waleed Mohamed 4241803 Seif Zaki 4231291
- 13. THANK'S