Today’s Objective
Week 3 introduces students to the performative aspects of AI, emphasizing how human and non-human agents interact in scientific practices. Drawing from Andrew Pickering's work, the class will explore the dynamic interplay between human intentionality and machine behavior, emphasizing the emergent nature of AI practices. By unpacking the 'mangle of practice,' students will gain a nuanced understanding of how science, agency, and time coalesce in AI.
Lecture Notes
Key Concepts
- Actor-Network Theory: A theoretical approach to understanding the relationship between human and non-human actors in a network (from Pickering's "The Mangle of Practice").
- Agency: The capacity of actors (human or non-human) to act in the world.
- Emergence: The process of complex patterns arising out of simple interactions.
To Read (Before Class)
Primary
- Andrew Pickering's "The Mangle of Practice: Time, Agency, and Science" (Preface & Chapter 1) The Mangle of Practice
Supplementary
- Bruno Latour's "Reassembling the Social" (Chapter 2) Reassembling the Social
Student Facilitation Reading
"Sorting Things Out: Classification and Its Consequences" by Geoffrey C. Bowker and Susan Leigh Star (Read & Present on Introduction and Chapter 1)Sorting Things Out
Summary: This reading explores the role of classification systems in shaping human and technological interaction. It offers insights into how categorization influences AI practices and connects with the theme of AI's performativity.
Critical Questions:
- How do classification systems impact the development and functioning of AI systems?
- What ethical or social issues may arise from AI classification practices?
- How does this reading relate to the concept of "mangle of practice" in AI?
- How do classification systems embody cultural or societal values?
- Can classification in AI be truly objective, or is it inherently biased?
To Watch
I’m not sure if this actually helps explain Actor Network Theory, but it was too odd not to share.
In Class Assignment
Actor-Network Mapping Exercise
- Objective: To understand the concept of Actor-Network Theory and The Mangle of Practice through hands-on experience.
- Method: Students will form small groups and select a current AI technology (e.g., a recommendation engine, self-driving cars, etc.). Using large sheets of paper and colored markers, each group will draw out the network or mangle associated with their chosen technology.
- Activity Details:
- Step 1: Identify all the actors (both human and non-human).
- Step 2: Draw lines to indicate the relationships or interactions between these actors.
- Step 3: Add notes to describe the nature of each relationship (e.g., influences, governs, disrupts).
- Wrap-up: Each group presents their actor-network map and discusses the emergent properties of their chosen system, referring to Latour's and Pickering's theories. You will answer one of the following questions:
‣
I’ve included the persona that might ask this question in parenthesis. The goal for answering these questions is to apply the theory.
- Regulatory Gaps (Legislator): Based on the network map, what regulatory measures are currently lacking or need revision to ensure equitable access and ethical use? (Legislator)
- Public Interest (Legislator): How do the different actors in this network contribute to or undermine the public interest? What legislative measures could address these issues?
- User Experience (Product Manager): What elements within this actor-network have the most significant impact on user experience, and how can we improve or iterate upon them?
- Resource Allocation (Product Manager): Given the complex interactions within the network, where should resources be allocated for maximum effectiveness and efficiency?
- Player Agency and Interaction (Game Designer): Within this network, what elements enhance or restrict player agency? How can we optimize the game environment to balance player freedom and intended game outcomes?
- Monetization vs. Ethical Gameplay (Game Designer): What are the ethical considerations in the network concerning player data, in-game purchases, and overall fairness? How can these be aligned with monetization strategies?
- Strategic Partnerships (Executive): Looking at the network, what strategic partnerships can be formed to augment company growth and to leverage technological capabilities?
- Long-Term Viability (Executive): How do the components and actors in the network contribute to or jeopardize the long-term sustainability and competitiveness of the business?
- Sustainability Concerns (Environmentalist) What elements within this actor-network have the most significant environmental impact, both positive and negative?
- Activism and Change (Environmentalist): How can this actor-network be leveraged to bring about environmental policy changes or promote sustainable practices?
Examples for the Mapping Exercise
‣
- Chatbots & Virtual Assistants
- Examples: Apple's Siri, Amazon's Alexa, Google Assistant
- Key Considerations: User interface, data collection, natural language processing
- Recommendation Engines
- Examples: Netflix's movie suggestions, Spotify's music playlists, Amazon's product recommendations
- Key Considerations: User preferences, data algorithms, business objectives
- Facial Recognition Systems
- Examples: Apple's Face ID, airport security systems
- Key Considerations: Privacy concerns, racial and gender biases, security measures
- Natural Language Processing Tools
- Examples: Google's BERT, OpenAI's GPT-4, Automated translation services
- Key Considerations: Language biases, data training, accuracy and limitations
- Autonomous Vehicles
- Examples: Tesla's Autopilot, Waymo's self-driving cars
- Key Considerations: Safety protocols, decision-making algorithms, legal frameworks
- Healthcare AI Systems
- Examples: IBM's Watson for Health, Google's DeepMind Health
- Key Considerations: Medical ethics, data accuracy, diagnostic efficiency
- Game-playing AI
- Examples: DeepMind's AlphaGo, OpenAI's Dota 2 bots
- Key Considerations: Machine learning techniques, human competition, strategic decision-making
- Sentiment Analysis Tools
- Examples: Social media monitoring tools, customer feedback analyzers
- Key Considerations: Emotional nuance, marketing strategies, public relations
- Financial Algorithms
- Examples: High-frequency trading bots, credit risk assessment algorithms
- Key Considerations: Market volatility, ethical trading, financial regulations
- Content Generation & Curation
- Examples: Automated news writing, social media feeds
- Key Considerations: Authenticity, clickbait risks, information ethics
‣
Human Actors
- Developers: The engineers who design and build the AI systems.
- Users: People who interact with the AI—ranging from everyday consumers to specialized professionals.
- Regulators: Individuals or groups responsible for setting rules and guidelines.
- Investors: Those who fund the development and deployment of these technologies.
- Ethicists: Professionals evaluating the ethical implications and considerations.
- Data Scientists: Those responsible for training models and handling the data that feeds into the AI.
- Media: Journalists and influencers who shape public opinion about the technology.
Non-Human Actors
- Algorithms: The mathematical rules that govern the behavior of the AI.
- Data Sets: The information used to train and/or run the AI.
- Hardware: The physical servers and devices where the AI is hosted or deployed.
- User Interfaces: The methods by which humans interact with the AI (e.g., screens, voice commands, etc.).
- APIs and Libraries: Code that connects the AI system to other systems.
- Network Infrastructure: Includes elements like internet bandwidth, cloud storage, and data centers.
- Sensors: In the case of autonomous vehicles or IoT devices, the sensors that gather real-time data.
- Electricity Supply: Energy source, often overlooked but crucial for operation.
Conceptual Elements
- Ethical Guidelines: Ethical considerations like fairness, accountability, transparency.
- Laws and Regulations: Specific legal frameworks that the technology has to comply with.
- Economic Factors: Market demands, revenue models, etc.
- Social Impact: Public opinion, potential for social change or disruption.
- Time: Both in terms of the historical development of the technology and real-time decision-making processes.
Environmental Actors
- Geographical Location: Data centers' physical locations and their environmental impact.
- Natural Resources: Consumed in the manufacturing and running of the technology.
- Carbon Footprint: Emissions related to the AI system’s lifecycle.
To Do
Read today’s materials before class & take notes on your reading
Come to class and participate in the readings and activities
‼️ Write a 200 word journal prompt on: What sort of scientific practices or procedures might influence the way we build and use AI today? Tell me how those considerations might, in turn, shape the AI technological development. Be sure to use the readings as background and include at least one quote or key passage from the reading.
Turn in all your assignments on Friday on eLearning