Please enable JavaScript.

Coggle requires JavaScript to display documents.

ITAIA1-22 AI ethics, AI ethics in ECE - Coggle Diagram

- - - - These resources can be bought or sold. They are available to everyone in the market and therefore provides no competitive advantage, but they are necessary but not sufficient by themselves to create capabilities.
      - Data
        
        2 types of data available to organisations
        
        Internal data
        
        External data
        
        To get a competitive advantage there must be integration between the internal and external data sources while at the same time ensuring that data is cleansed, processed and distributed as required, within the applicable boundaries and/or laws and regulations.
      - Technology
        
        Using AI and leveraging the large amounts of data it uses requires a technological infrastructure that can support the data storage and processing requirements. Many organisations are moving to cloud-based infrastructure to reduce the cost associated with AI infrastructure. The lack of a technological infrastructure in an organization is often highlighted as one of the main barriers to AI adoption.
      - Basic Resources
        
        Time and financial resources are required to allow AI initiatives to deliver the expected outcomes. Therefore there must be adequate financial resources invested for the outcomes to materialize. This allocation of financial resources will obviously compete with other priority items that also need financial resourcing.
    - - The knowledge, skills, experience, leadership, problem-solving skills, communication and collaboration competencies of the human workforce of an organization is what is assessed to measure the human capital of that organization.
      - Technical skills
        
        These are the skills required to deal with the implementation and realization of the AI algorithms, its infrastructure as well as the applications. It also refers to the skills required to train the AI and the skills required to ensure that business managers understand how the AI functions but in a non-technical manner.
      - Business Skills
        
        This refers to the managerial skills to know where and how to deploy the AI technologies, the potential of these technologies and the potential impact and ramifications of the use of these technologies, such as with regards to privacy, safety etc.
    - - These resources are difficult to replicate and are very important especially when markets or environments are uncertain and volatile. These resources are also not similar across organisations and are unique to most.
      - Inter-departmental coordination
        
        This is a key enabler of innovation and creativity and is regarded as a cornerstone of success for multi-disciplinary projects. Enhanced coordination between departments ensure that these organisations will become more agile and adaptable in deploying AI applications and technologies.
      - Organisational change capacity
        
        A key success factor in digital transformation is the ability to start and complete the execution of plans. This capacity is focused on the ability to deal with the potential problems that might arise as a result of transitioning from the old to the new processes. This capacity deals with, inter alia, the breaking of the status quo, introducing new practices, new values and new structures. It is also important to note that the ability to overcome resistance to change is an important component of this capacity.
      - Risk proclivity
        
        Being pro-active and adopting a risk-taking stance has been shown to leading to a reaping of benefits long before competitors do, and this is also the case when adopting AI into organisations and as part of the organizational operations. This pro-active and risk-taking stance is also commonly associated with higher levels of innovation and market leadership.
- - - - Both human agency and machine agency involve the capacity to perform actions and tasks, and to affect other agents and the environment. 
      - Both human agency and machine agency depend on the design, programming, and data that shape their goals, values, beliefs, and motivations.
      - Both human agency and machine agency can interact and collaborate with each other and create synergies and outcomes that are greater than the sum of their parts.
    - - Some possible differences between human agency and machine agency are:
        
        Human agency is based on the capacity to make choices and decisions, and to act upon them, based on one’s goals, values, beliefs, and motivations. 
        
        Human agency is influenced by various factors, such as emotions, intuitions, morals, and social norms, that can affect one’s judgment and behavior.
        
        Human agency can be exercised in various domains and contexts, and can adapt to changing situations and environments. 
        
        Human agency can be creative and innovative and can generate novel and original ideas and solutions. Human agency can be ethical and responsible and can consider the consequences and impacts of one’s actions and decisions.
- - - - Could not learn from data,
      - Couldn’t adapt to new situations,
      - Required frequent manual updating.
  - - - Machine learning (ML) allows systems to learn patterns from data instead of being explicitly programmed.
      - Advantage over symbolic AI:
        
        Can improve performance as more data becomes available.
        
        Less reliant on human-provided rules.
  - - - Image recognition,
      - Natural language processing,
      - Speech recognition.
  - - - Lacks moral understanding,
      - Cannot be held morally responsible,
      - Should not be treated like a person.
  - - - Diverse and representative data,
      - Bias detection and correction methods.
- - - - Text
      - Images
      - videos, and audio
  - - - Natural language generation
      - Speech recognition
      - Virtual agents
      - Decision management
      - Biometrics
      - Machine learning
      - Robotic process automation
      - Peer-to-peer network
      - Deep learning platforms
      - AL-optimized hardware
    - - AI-generated Art
      - Generative AI Models
      - Democratization of AI technology
      - No-Code/Low-Code Platforms
      - AI-Powered Apps
      - AI Ethics and Fairness
      - Bias Removal
      - Explainable AI
      - Transparency
      - Cybersecurity
    - - Edge Computing
      - Collaboration between Humans and AI
    - - Health & Medical
      - Auto & Road Safety
      - Agriculture
      - Entertainment
      - Service Industries
      - Military
      - Communications
      - Research and Development
      - Production
      - Logistics
      - Shipping
      - Safety and Security
    - - Another area that will be affected greatly, are the different data types. That is because AI can process and learn from the various types of data. Depending on which area above we are looking at, different data types will be affected more or less. Here is a summary of the data types that may be affected the most:
      - Text
      - Images
      - Audio
- - - - Doing Good with AI
        
        6 Guidelines
        
        The Montreal Declaration says AI should promote the well-being of “all sentient creatures.”
        
        The IEEE principles prioritize human well-being in the design of all AI systems.
        
        The UK House of Lords (AIUK) and Asilomar Principles describe this as developing AI “for the common good.”
        
        The Partnership on AI wants AI to “empower as many people as possible.”
        
        The EGE (European Group on Ethics) expands this to include human dignity and sustainability, arguing AI should help protect life and the environment for future generations.
      - Promoting Well-being
      - Preserving Dignity
      - Sustaining the Planet
    - - Avoiding Harm
        
        4 Guidelines
        
        Most guidelines highlight the risk of violating personal privacy.
        
        The Asilomar Principles warn against AI weapons or systems that improve themselves uncontrollably.
        
        The Partnership on AI emphasizes “secure constraints” to prevent AI from doing harm.
        
        The IEEE and Montreal Declaration talk about developers taking responsibility for the risks their innovations may pose.
      - Security
      - Privacy
      - “Capability Caution”
    - - Keeping Human Control
        
        4 Guigelines
        
        The Montreal Declaration says AI should support not undermine human autonomy.
        
        The EGE warns that machines should not limit people’s ability to set their own values.
        
        The AIUK states that no AI should be given the power to hurt or deceive people.
        
        The Asilomar Principles emphasize that humans must choose how and when to let AI make decisions.
        
        meta-autonomy
      - The Power to “Decide to Decide”
    - - Ensuring Fairness and Solidarity
        
        The Montreal Declaration calls for eliminating all forms of discrimination in AI systems.
        
        Asilomar emphasizes “shared benefit” and “shared prosperity.”
        
        The EGE focuses on “justice, equity, and solidarity,” warning against biased algorithms and highlighting the need to protect systems like social insurance and healthcare.
      - Promoting Prosperity
      - Preserving Solidarity
      - Avoiding Unfairness
    - - This principle is uniquely important in AI because it adds a layer of clarity and accountability to how AI systems work and who is responsible for their actions.
      - Making AI Understandable and Accountable
      - Intelligibility: Can people understand how the AI system works?
        Accountability: Can we identify who is responsible for what the system does?
      - Explicability enables the other four principles. If we don’t understand how an AI system works:
        
        We can’t be sure it’s doing good (beneficence),
        
        We can’t prevent harm (nonmaleficence),
        
        We can’t know when to trust it or take over (autonomy),
        
        And we can’t hold anyone accountable if it’s unfair (justice).
- - - - breaches
      - profiling
      - misuse
  - - - social robots
      - emotion recognition technology
      - BUT
        
        risk undermining relational learning
        
        fostering overreliance
        
        manipulation
        
        loss of autonomy
  - - - perpetuates systemic inequities
        
        affecting marginalised communities
- - - - Transparency
      - Data minimization
      - Cultural inclisitivity