AI legislation Flashcards
(21 cards)
AI legislation
→ Need of ethically-inspired and ethically- driven policies concerning AI, it’s regulation and governance
AI legislation
→ Governance
Distinguish between:
→ Governance BY algorithms
- Decision through algorithms
→ Governance OF algorithms
- actions regulating algorithms
AI legislation
→ Governance
- Governance by algorithms
→ Controversial
- diverging interests
- application range
- forms
→ Shape sectors such as
- Public administration
- Media
- Digital work
- DIgital freedom
AI legislation
→ Search engines
→ Digital universe is a public space
- Freedom of expression
- Discussion as an integration-driving force
→ Search engines create a visibility hierarchy of available information
- Google’s algorithm
- Amazon algorithm
- Automated moderation of social media
- Automation of PA
AI legislation
→ Search engines
- Google’s algorithm
→ PageRank
→ Criteria
- prestige of info (#mentions)
- audience (# visits)
- closeness and affinity to the user (profiling)
- fastness (newness)
→ Not completely understood because of industrial secret
AI legislation
→ Search engines
- Amazon’s algorithm
→ Purchase history of users that have bought the same book
→ Analyze suggestion and recommendations
→ Relies on
- Data mining
- Affinity analysis
- Market basket analysis
AI legislation
→ Search engines
- Automatized moderation on social media
→ Ensure contents conform to laws and the platform’s specific rules
- not achievable only by humans
- to cut costs automatized system to identify
▪ hate speech
▪ disinformation
▪ black illegal
▪ copyright violation - after algorithm identify potential problematic contents human verify
▪ secrecy on criteria and technology
▪ Global Internet Forum of Counter Terrorism (GIFCT)
Meta Google Microsoft and X shared a database - ethical questions
▪ respect of dignity - political questions
▪ risks for democracy
AI legislation
→ Search engines
- Automatized on PA
→ More efficient by
- rationalizing bureaucracy
- improving information research
- improving the choice between available options
→ Cons:
- new forms of surveillance
- people profiling by institutions
AI legislation
→ Algorithm governance
→ Technology mirrors and reshapes society
→ To solve ethical and political issues by AI
→ From principles to laws
AI legislation
→ Algorithm governance
- Technology mirrors and reshapes society
Technology artifacts contribute to govern society and social interactions
AI legislation
→ Algorithm governance
- To solve ethical and political issues by AI
Assume fundamental ethical principles →
→ Translate into policies concerning AI →
→ Inspire a legal regulation of AI
Ethics → Policy → Legislation
AI legislation
→ Algorithm governance
- From principles to laws
→ Deeper supervision of algorithms
→ Automatized processes should not replace human intervention
- only assistance in decision making
- GRDP already forbids decision-making producing legal consequences if is only by a fully automated treatment of data
→ information and decision asymmetry between users and developers/platform
- necessity to expand users protection
▪ transparency requirements
▪ prohibition to use any program leading to any form of discrimination
AI legislation
→ Examples
- Private codes
- Public codes
AI legislation
→ Examples
- Private codes
Professional deontological codes
→Two main characteristics
- personal integrity
- personal honesty
→ Guidelines
- to make decisions
- to educate public and professionals about the obligation that come with their profession
→ Examples
- ACM
- IEEE
AI legislation
→ Examples
- Private codes
ACM
Association for Computing Machine
→ World’s biggest association of IT
→ Code of Ethics & Professional conduct
- 4 Sections
▪ Fundamental ethical principles
▪ Specific professional responsibilities
▪ principles concerning those charged with executive or leadership roles
▪ adoptions of code and situations when it is trespassed (sanctions)
AI legislation
→ Examples
- Private codes
IEEE
Institute of Electrical & Electronics Engineers
→ Ethical and deontological code
- Developers should
▪ take into account ethical and social concerns
▪ how to improve society (aspirational)
AI legislation
→ Examples
- Public codes
→ High level expert group on AI (EU)
→ UN interim report - Governing AI for humanity (2023/12)
→ EC white books on AI (2020)
AI legislation
→ Examples
- Public codes
→ High level expert group on AI
→ multilevel approach
- mix of ethical and social
- main interests
▪ scientific
▪ industrial
▪ financial
▪ commercial
▪ political
→ equal distribution of benefits and risks
→ central notion
- ethically oriented AI
▪ flourish sector
▪ attract funds and talents
→ trustworthy AI
- three requirements
▪ Robust systems
▪ Ethical values
▪ Ethical principles
→ Practical context
- main goals
▪ meaningful human control
▪ technical soundness and safety
▪ privacy
▪ transparency
▪ diversity
▪ social and environmental well-being
▪ Accountability
AI legislation
→ Examples
- Public codes
→ UN interim report
→ Governing AI for humanity (2023/12)
- Principles:
▪ AI should be governed inclusively
▪ AI should be governed in the public interest
▪ AI governance should be built in step with data governance and the promotion of data commons
▪ AI governance must be universal, networked and rooted in adoption multi-stakeholder collaboration
▪ AI governance should be anchored in the UN charter, international Human Right Law, and other agreed international commintments
AI legislation
→ Examples
- Public codes
→ EC white books on AI
→ Human operators to intervene and surveill over decision guided by AI
→ Safe infrastructure
→ Guaranteed privacy
→ Avoid opacity
→ Diversity of data used in AI
→ Goal of social and environmental welfare
→ Clear roles and responsibilities
AI legislation
→ Governing AI by codes and rules
→ Principles and recommendations drawn to
- outline professional responsibilities
- influence future development in the IT industries
→ 3 problems
- Ethical washing
▪ only to calm public opinion
▪ codes are vague
▪ sanctions does not work - Difficult to apply
▪ not specify how to behave
▪ impossible to enact what is found in codes - Pluralism of values and moral autonomy
▪ Ethics requires capacity of judgement not submission to rules
▪ Values diverge
▪ Limiting autonomy
