For my 10-week thesis project at the Copenhagen Institute of Interaction Design, I combined my backgrounds in philosophy and design to explore the ethics of autonomous machines and autonomous decision making.
The result is the Machine Ethics Toolkit. A DIY workshop for artificial intelligence and machine learning companies.
Machine ethics debates what decision an autonomous machine should make when presented with two unfavorable choices. With this toolkit, you can facilitate a workshop to learn philosophical moral theories, ethical decision-making framework, and vocabularies to talk about ethics. This toolkit aims to encourage companies to think critically about their work and to integrate ethical considerations into their workflow.
The toolkit consists of a playbook for the facilitator and worksheets for the participants
01
Autonomous machines are entering into every facet of our lives. They are increasingly dealing with difficult human decisions. Google’s Waymo driverless car technology has just been approved for public roads. The aging demographic crisis is propelling the development and acceptance of elderly care robots. And we now have invisible algorithms that make decisions on employment and criminal sentencing. As designers, engineers, and creators of machine learning and artificial intelligence technology, we need to include ethical considerations into our product and services.
A potential future scenario with Alexa
02
With this toolkit, you can facilitate a workshop on machine ethics for a group of 6-8 people. The workshop takes a total of 150 - 180 minutes and consists of two parts. The first part is to learn about ethical approaches. The second part is to work through the ethical decision-making framework. There are a total of 6 activities in this workshop.
Workshop activities overview
The first part of the workshop is aimed at familiarizing the participants with the concepts and language of ethics.
Activity 1: We map out how each participant feels about the future of autonomous machines on a 2 x 2 grid.
Activity 2: I introduce two classic ethical dilemmas to the participants to get them use to the idea of deciding on the most preferable decision.
Activity 3: I divide up the participants into teams of two and introduce the trolley problem and the three classic ethical theories: consequentialism, duty, and virtue ethics.
The Trolley Problem is used to explain the different ethical frameworks
In the second part of the workshop, we looked at a specific scenario and the ethical decision making framework.
Activity 4: Understand the scenario of an autonomous carebot assisting a caretaker in giving an elderly a bath.
Activity 5: The participants work through 5 worksheets that represented the ethical decision-making framework.
Activity 6: We discussed the most ethical decision and reflect on the workshop.
The main scenario card
The steps in the Ethical Decision Making Framework
03
Participants responded positively to the workshop. Those who work in the area of machine learning wanted to see this ethical framework applied to their professional work. Participants to took away three learnings from this workshop.
1. An understanding of the three major ethical theories (Utilitarian, Duty, and Virtue)
2. How to use a framework for making ethical decisions
3. Reflect critically on their professional work in machine learning
04
The topic of machine ethics is vast and complex. I realized early on that I needed to supplement the Human Centered Design Process. Asking non-experts what they thought of machine ethics lead to a lot of blank stares. My initial research phase was composed of secondary desktop research (Click here to see a list of my references) and attending the Machine Ethics Summit during the 2017 Copenhagen TechFest.
I gave a lightning talk at the Machine Ethics summit on critical design projects that inspire me
Group brainstorm around the topic of ethical machines
The most challenge task that I need to figure out is how to reduce the complexity of the topic. After some discussion with philosophers and designers, I decided to direct my focus on the mundane and look at the ethics of automation in everyday situations. Two insights that came out of the research phase is that;
1. Ethics is not about right or wrong. It’s about the most preferable decision.
2. We all have a basic and innate understanding of ethics. But ethics is difficult to talk about without scenarios, vocabulary, and framework.
A look at my chosen topic through a philosophical lens and the potential questions for exploration
05
Initially, I wanted to create a tangible experience for machine ethics. My early sketches and explorations are all tangible experiences. However, none of these experiences taught people about ethics. I felt stuck. In order to move forward, I created two sacrificial prototypes. The first prototype is a video where I had a student narrate the “thoughts” of autonomous machines while they completed various task. In the second prototype, I role-played as an “Alexa” in the studio where anyone can ask my to perform Alexa tasks. As much as I enjoyed these two prototypes, they failed in exploring the topics of machine ethics. They taught me what not to do.
I roleplayed as an Alexa to try and empathize with machine intelligence
A video where a student is asked to narrate and humanize the autonomous decision making process of machines
I again turned to the experts in this field and started to have more conversations around Machine Ethics. I realized that a simple interactive object cannot bring out all of the complexities of Machine Ethics. I had to think more broadly about what the deliverable of this project could be. Because I've had so much success talking with experts about this topic, I decided to think about a workshop as the final deliverable.
An overview of the people who I spoke to for this project
06
Once I established that I wanted to create a workshop toolkit around my topic, I need to figure out how to create an engaging workshop. I set a goal of creating one workshop every week. I ran five workshops in five weeks. During these prototype workshop, I explored different workshop techniques. In one workshop, I was inspired by Dungeons and Dragon to incorporate roleplaying and group decision-making. I had the participants act out an ethical dilemma. The various prototype workshops helped me to understand how to craft the right scenario and how to design engaging activities. I created the final toolkit using the learnings from the five workshops.
An overview of the five prototype workshops and the question that I tried to answer with each.
Week 2 workshop on smart objects
Week 7 workshop on scenario building
07
First, I learned how to scope a huge and difficult topic. Second, I learned how to design an engaging workshop. To help me scoop my project, I had to consistently make low fidelity prototype. These prototypes can be as complicated as an experience or as simple as a photo. I’ve also learned to let go of my high expecations and accept that a 10-week project is not going to solve machine ethics.
Final project exhibition