Published on

Practical AI Ethics course notes - Ethics toolkit


Lesson 3 - AI Ethics toolkit

Ethical lenses/frameworks

Deontological Ethics

  • Focus on rights, principles and dutires
  • Principles include: justice,. fairness, consistency, transaprency
  • There’s a strong notion of “Right” and “Wrong”
  • Examples
    • The Golden Rule ??
    • Kant’s formula of Humanity - I should always treat other persons as ends in themselves, never just means to an end
    • Justice approach - Whcih option treats people equally or proportionately
  • Typical questions to think through
    • What rights of other & duties must be respected?
    • How might the automomy and dignity of stakeholders be respected?

Consequentialist ethics

  • Focus on consequences of applied technology
  • Examples
    • Utilitarian approach: Which option will produce the most good and the least harm?
    • Common good approach: Whcih option best serves the commnuity rather than the individual?

Virtue ethics

  • Focus on the actor/character traits
  • Development of practical wisdom
  • Examples
    • The virtue approach: Which option leads me to act as the sort of person I want to be
  • Typical questions to think through
    • What design habits are we embodying? Are they the habits of excellent designers? Who do we look up to?
    • Will this design/project incentivize any vicious habits in users/stakeholders?
    • How confident are we that we will individually and, as a team/organisation, be proud to have our name associated with the project one day?


Western philosophies, there a plenty other ethical lenses

Ethical toolkit for engineers/designers

1. Ethical Risk Sweeping

  • Institute regularly scheduled ethical risk-sweeping practices
  • Treat in the same way you do cybersecurity - what is not found is good news, but that doesnt meant you should stop doing it

2. Ethical Pre-Mortems & Post mortems

  • Team/Project pre-mortem should ask ⭐
    • How could this project be a fail for ethical reasons?
    • What blind spot would lead us into it?
    • Why would we fail to act?
    • What systems/Processes/Check can we put in place to reduce failure rish
  • Team/Project post-mortem should ask

3. Case based analysis

  • Identify similiar paradigms or cases that mirror the present case
  • Evaluate choices made and outcomes betweencases

4. Remembering the Ethical benefits of Creative work

  • Why are we doijng this, and for what good ends
  • Will society, the world, our customers really be better off with this tech than without it/

5. Closing the loop / ethical feedback and iteration

  • Identify feedback channels that will deliver reliable data on ethical impact
  • Integrate the process with quality management / user support
  • Develop formal procedures and chains of responsability for ethical iterations

Personal notess / Riffin’ away

  • Talking to engineers and tech leads around ethics in their work, the biggest blocker practically seem to be when the ethical questions and work is separated from the project (e.g done in a similiar way as a security audit); The ethicals concerns usually come too late to the party, and delivery teams are avoiding addressing the outcomes to reach their targets

  • The ideal framework imho would be embedded in the project development, involving team members and stakeholders from the project start. Particularly interested in the pre-morterm process: questions are dead simple yet powerful in unravelling some ethical issue at project scoping, it should be relatively easy to implement in a scoping/solution design process

  • Ethics is most definitelty some of the hardest questions and can produce vague answers, uncertainty and could potentially shatter a team’s confidence; while I think all team members should understand a company’s values and ethical framework, I can also see the challenge in making such questions a team concern and priority

  • One famous adage I learned from my years in software that resonates true to me every single day, most particularly when attempting to improve the status quo and progress practices and processes that are tedious and cognitively demanding, is to “make the change easy, then make the easy change”

    • Ethics, similiar to security can very much be seen as a friction point and a slow down to the design process and prospective delivery outcome; What are ways to address this in the most frictionless way
    • The premortem solution seem like a good start in at least raising the biggest risks and in building the in-house ethical framework of a company
    • Also interested in “Ethics as a service” and how such service could interogate a project with expertise while be embedded into the regular project delivery
    • AI embodying the company’s values/ethical framework and understanding a project ideals could be trained to prompt ethical questions and point at risks (it may not sound super ethical coming from AI, but I believe that AI is after all a form of automation, which is most often the fabric of making change easy)
Let's stay in touch!
Get the latest posts delivered right to your inbox