What do you think?
Rate this book
Ethics for Robots
Ethics for Robots describes and defends a method for designing and evaluating ethics algorithms for autonomous machines, such as self-driving cars and search and rescue drones. Derek Leben argues that such algorithms should be evaluated by how effectively they accomplish the problem of cooperation among self-interested organisms, and therefore, rather than simulating the psychological systems that have evolved to solve this problem, engineers should be tackling the problem itself, taking relevant lessons from our moral psychology.
Leben draws on the moral theory of John Rawls, arguing that normative moral theories are attempts to develop optimal solutions to the problem of cooperation. He claims that Rawlsian Contractarianism leads to the 'Maximin' principle - the action that maximizes the minimum value - and that the Maximin principle is the most effective solution to the problem of cooperation. He contrasts the Maximin principle with other principles and shows how they can often produce non-cooperative results.
Using real-world examples - such as an autonomous vehicle facing a situation where every action results in harm, home care machines, and autonomous weapons systems - Leben contrasts Rawlsian algorithms with alternatives derived from utilitarianism and natural rights libertarianism.
Including chapter summaries and a glossary of technical terms, Ethics for Robots is essential reading for philosophers, engineers, computer scientists, and cognitive scientists working on the problem of ethics for autonomous systems.
Leben draws on the moral theory of John Rawls, arguing that normative moral theories are attempts to develop optimal solutions to the problem of cooperation. He claims that Rawlsian Contractarianism leads to the 'Maximin' principle - the action that maximizes the minimum value - and that the Maximin principle is the most effective solution to the problem of cooperation. He contrasts the Maximin principle with other principles and shows how they can often produce non-cooperative results.
Using real-world examples - such as an autonomous vehicle facing a situation where every action results in harm, home care machines, and autonomous weapons systems - Leben contrasts Rawlsian algorithms with alternatives derived from utilitarianism and natural rights libertarianism.
Including chapter summaries and a glossary of technical terms, Ethics for Robots is essential reading for philosophers, engineers, computer scientists, and cognitive scientists working on the problem of ethics for autonomous systems.
- GenresPhilosophyNonfiction
166 pages, Hardcover
Published July 23, 2018
2 people are currently reading
111 people want to read
Ratings & Reviews
Friends & Following
Create a free account to discover what your friends think of this book!
Community Reviews
Displaying 1 - 5 of 5 reviews
January 3, 2020
Ethics for Robots: How to Design a Moral Algorithm by Derek Leben
Robot Ethics is an attempt to tackle a popular topic of Ethics in AI by looking for a suitable moral theory first.
Ethics in AI is a popular topic, but surprisingly little work is dedicated to the ethics side of the problem. In his debut book, Leben starts by tackling the ethical side of the equation first. He starts the book with the embraces of evolutionary hypothesis - the same conclusion Darwin puts forward in The Descent of Man (1871) - that says that “moral judgments are the product of evolutionary pressure for cooperation in self-interested organisms” or as Leben puts it, “moral rules are adaptive solutions to cooperation problems.”
An evolutionary explanation of morality is a popular stance, and not just among sociobiologists. However, Leben glosses over its arguments and moves on to game theory. He advises us to look beyond the Nash Equilibrium solutions - points where all players have maximized their payoffs while playing their best game - and go for Pareto-improvements instead. Pareto-improvements are alternative points to Nash Equilibria where already good results can be improved “to make at least one person better off without making any other individual worse off" and constitute the core of Leben’s approach in the book.
Leben then proceeds to evaluate historically influential moral theories such as utilitarianism, libertarianism, Kantian ethics, and virtue ethics. In a brief but well-reasoned stream of arguments, Leben succeeds in showing that all four theories fail to be internally consistent in promoting instrumental values. For Leben, moral theories are about instrumental values and internal consistency refers to a game-theoretical proof that theory will yield a Pareto-improvement for every given set of circumstances. This test is especially important if we plan on developing algorithms according to moral theories.
The moral theory suitable for machines
According to Leben, the theory that “necessarily ensure cooperative behavior in every situation” is Contractarianism. The theory has its roots in the writings of Thomas Hobbes (1679) and Jean-Jacque Rousseau (1778) and, more recently, of David Gauthier and John Rawls (2002). For Leben, the essential concepts in Contractarianism are Rawls’ concepts of original position and the veil of ignorance. The two concepts comprise a thought experiment where we are asked to decide how to act without knowing what the circumstances are. In other words, we are asked to agree on instrumental values by assuming we can end up being anyone. Rawls claims that in such a position, we would all agree to make the worst-off person as best-off as we can because we can end up being that person. The procedure of making the worst-off person as best-off as possible is called the Maximin principle, and Leben argues that it should be the critical component of an ethical algorithm.
The rest of the book is an attempt to devise such an algorithm. With a style focused on brevity, Leben applies Maximin principle, compatible versions and derivatives of it, such as Leximin, to some of the ethical dilemmas involving machines. From self-driving cars and models for predicting recidivism (COMPAS) to rescue operations, bioethics, and autonomous weapons, Leben does an excellent job in framing the conversation by starting with the ethical principle first.
In the chapter about self-driving cars, the ethics-first approach forces Leben to consider both the likelihood of a collision and the estimated damage to each agent. That allows the vehicle to calculate the risk of harm (h) score for each decision. With such a score at hand, it is easy to test the outcomes using various moral theories. The utilitarian vehicle would aim to minimize the total sum of h for all collisions, self-interested cars to reduce the damage to the car and the passenger, and Contractarian vehicle would try to avoid the path leading to the highest value of h.
This approach is not without problems. First, as a true Contractarian, the autonomous car running Leben’s algorithm would have to take into consideration only the loss of primary goods. According to Rawls, these are the goods that are supposed to be desirable by all, such as intelligence, imagination, health, rights, liberties, income and wealth, the social bases of self-respect. Such goods are hard to quantify, let alone predict how each collision would affect each of them. Also, a vehicle with the Contractarian algorithm may be less appealing to passengers, who would, in spite of their (publicly) stated moral preferences, pick a car that will protect them at all costs.
No Trolley Problem debate dilemma is complete without bringing up the Moral Machine game, one of MIT’s longest-running experiments on moral preferences. In the game, a player is given a choice between two paths, both deadly but for different kinds of people or animals. The player gets to pick who he/she would want to run over and who to spare - young, old, homeless, affluent, pregnant, dogs, cats, etc. Leben is very critical of the Moral Machine because of the use of social properties and believes that “[the] experiment is mistakenly testing people’s discriminatory biases rather than their moral judgments.”
Leben’s work is far from a definite answer to many of the questions raised in the debate on ethics and machines. Because of its brevity, that leaves many questions unanswered, the book will likely leave both ethicists and the software developers complaining. However, the book is different enough from most contemporary writing on the topic of ethics and AI, and despite being short and directional, it is a laudable work.
Derek Leben is the Assistant Professor of Philosophy at the University of Pittsburgh, Johnstown, whose work lies at the intersection of ethics, cognitive science, and technology.
Robot Ethics is an attempt to tackle a popular topic of Ethics in AI by looking for a suitable moral theory first.
Ethics in AI is a popular topic, but surprisingly little work is dedicated to the ethics side of the problem. In his debut book, Leben starts by tackling the ethical side of the equation first. He starts the book with the embraces of evolutionary hypothesis - the same conclusion Darwin puts forward in The Descent of Man (1871) - that says that “moral judgments are the product of evolutionary pressure for cooperation in self-interested organisms” or as Leben puts it, “moral rules are adaptive solutions to cooperation problems.”
An evolutionary explanation of morality is a popular stance, and not just among sociobiologists. However, Leben glosses over its arguments and moves on to game theory. He advises us to look beyond the Nash Equilibrium solutions - points where all players have maximized their payoffs while playing their best game - and go for Pareto-improvements instead. Pareto-improvements are alternative points to Nash Equilibria where already good results can be improved “to make at least one person better off without making any other individual worse off" and constitute the core of Leben’s approach in the book.
Leben then proceeds to evaluate historically influential moral theories such as utilitarianism, libertarianism, Kantian ethics, and virtue ethics. In a brief but well-reasoned stream of arguments, Leben succeeds in showing that all four theories fail to be internally consistent in promoting instrumental values. For Leben, moral theories are about instrumental values and internal consistency refers to a game-theoretical proof that theory will yield a Pareto-improvement for every given set of circumstances. This test is especially important if we plan on developing algorithms according to moral theories.
The moral theory suitable for machines
According to Leben, the theory that “necessarily ensure cooperative behavior in every situation” is Contractarianism. The theory has its roots in the writings of Thomas Hobbes (1679) and Jean-Jacque Rousseau (1778) and, more recently, of David Gauthier and John Rawls (2002). For Leben, the essential concepts in Contractarianism are Rawls’ concepts of original position and the veil of ignorance. The two concepts comprise a thought experiment where we are asked to decide how to act without knowing what the circumstances are. In other words, we are asked to agree on instrumental values by assuming we can end up being anyone. Rawls claims that in such a position, we would all agree to make the worst-off person as best-off as we can because we can end up being that person. The procedure of making the worst-off person as best-off as possible is called the Maximin principle, and Leben argues that it should be the critical component of an ethical algorithm.
The rest of the book is an attempt to devise such an algorithm. With a style focused on brevity, Leben applies Maximin principle, compatible versions and derivatives of it, such as Leximin, to some of the ethical dilemmas involving machines. From self-driving cars and models for predicting recidivism (COMPAS) to rescue operations, bioethics, and autonomous weapons, Leben does an excellent job in framing the conversation by starting with the ethical principle first.
In the chapter about self-driving cars, the ethics-first approach forces Leben to consider both the likelihood of a collision and the estimated damage to each agent. That allows the vehicle to calculate the risk of harm (h) score for each decision. With such a score at hand, it is easy to test the outcomes using various moral theories. The utilitarian vehicle would aim to minimize the total sum of h for all collisions, self-interested cars to reduce the damage to the car and the passenger, and Contractarian vehicle would try to avoid the path leading to the highest value of h.
This approach is not without problems. First, as a true Contractarian, the autonomous car running Leben’s algorithm would have to take into consideration only the loss of primary goods. According to Rawls, these are the goods that are supposed to be desirable by all, such as intelligence, imagination, health, rights, liberties, income and wealth, the social bases of self-respect. Such goods are hard to quantify, let alone predict how each collision would affect each of them. Also, a vehicle with the Contractarian algorithm may be less appealing to passengers, who would, in spite of their (publicly) stated moral preferences, pick a car that will protect them at all costs.
No Trolley Problem debate dilemma is complete without bringing up the Moral Machine game, one of MIT’s longest-running experiments on moral preferences. In the game, a player is given a choice between two paths, both deadly but for different kinds of people or animals. The player gets to pick who he/she would want to run over and who to spare - young, old, homeless, affluent, pregnant, dogs, cats, etc. Leben is very critical of the Moral Machine because of the use of social properties and believes that “[the] experiment is mistakenly testing people’s discriminatory biases rather than their moral judgments.”
Leben’s work is far from a definite answer to many of the questions raised in the debate on ethics and machines. Because of its brevity, that leaves many questions unanswered, the book will likely leave both ethicists and the software developers complaining. However, the book is different enough from most contemporary writing on the topic of ethics and AI, and despite being short and directional, it is a laudable work.
Derek Leben is the Assistant Professor of Philosophy at the University of Pittsburgh, Johnstown, whose work lies at the intersection of ethics, cognitive science, and technology.
September 1, 2019
It is easy to find things to disagree with in this book, but only because Leben forces himself to lay out the practical implications of his theory like I've seen few others do. When the rubber hits the road, there's no room for hand-waving and abstract arguments. This book is almost clinically cleansed from such ungrounded "airy" arguments.
I was left unconvinced by the contractarian view that Leben promoted (I find the utilitarianism more attractive), this book was great to reflect on.
A few basics I found unconvincing was (1) Leben asserts that it follows from the evolutionary origin of our moral intuitions, that the "right" choice is to respect them. I think we ought to aim higher. (2) The book contains many assertions about what people in the contractarian "original position" would agree to, including retribution for crimes and the maximin principle. If we endowed people with a bit of content knowledge and riddance of biases and statistical illiteracy, I think they would come to other conclusions.
I was left unconvinced by the contractarian view that Leben promoted (I find the utilitarianism more attractive), this book was great to reflect on.
A few basics I found unconvincing was (1) Leben asserts that it follows from the evolutionary origin of our moral intuitions, that the "right" choice is to respect them. I think we ought to aim higher. (2) The book contains many assertions about what people in the contractarian "original position" would agree to, including retribution for crimes and the maximin principle. If we endowed people with a bit of content knowledge and riddance of biases and statistical illiteracy, I think they would come to other conclusions.
September 3, 2019
This is a very clearly written book that addresses many ethical issues in a straightforward and logical way. It’s also reasonably short, which is an excellent quality for a book that hopes to have a positive impact on the world.
I’m uncertain that I fully agree with the “lexi-min” principle put forward here, but I very much appreciate the precision of argumentation, and it seems like a plausible starting place for many ethical questions.
I am skeptical that ethical principles this abstract can be implemented in the “automated ethics engine” Leben describes for at least the foreseeable future. However, I do think it’s a good framework for making engineering decisions and trade-offs in many areas, so I’d heartily recommend the book to all software engineers.
I’d also recommend the interview Leben did on Sean Carrol’s “Mindscape” podcast as an introduction or supplement to this book.
I’m uncertain that I fully agree with the “lexi-min” principle put forward here, but I very much appreciate the precision of argumentation, and it seems like a plausible starting place for many ethical questions.
I am skeptical that ethical principles this abstract can be implemented in the “automated ethics engine” Leben describes for at least the foreseeable future. However, I do think it’s a good framework for making engineering decisions and trade-offs in many areas, so I’d heartily recommend the book to all software engineers.
I’d also recommend the interview Leben did on Sean Carrol’s “Mindscape” podcast as an introduction or supplement to this book.
May 26, 2019
Not many can become a Henry Ford, or even an Elon Musk. But you can still make robots, even if you won't change the World. Even if you are so poor you can't make robots, or you don't know much about mechanics, you can still write software. What can you do when you are too intellectually limited to even do the programming? Well, you can teach others what they should do with their robots. Even be awarded a tenure as prize for... not enough brain to do math and physics?
June 26, 2025
I lost interest.
Displaying 1 - 5 of 5 reviews