Economics, Law and Ethics
All recordings are available on Moodle.
Introduction
Engineering is not just about writing clever code, but about creating whole systems that do useful work, and deploying them in ways that are economically sustainable and ethical. It extends from pure mathematics, semiconductor physics and electrical engineering, through our core computer science and systems engineering topics, through to economics, law and policy.
Economic arguments appear in many areas of computer science. Economics deals with mechanisms whereby global equilibria emerge from the local behaviour of a number of selfish agents. Economic arguments and techniques are used by computer scientists to tackle problems from network congestion, through resource allocation in distributed systems, to security. As systems become ever larger, and involve ever-more diverse stakeholders, these techniques become steadily more important. Some modern constructs, such as blockchains, cannot be analysed sensibly without paying attention to the incentives faced by participants.
Market mechanisms often work, but they can also fail, or yield an equilibrium far from the social optimum. This is common in our field because of the tendency of many information markets to monopoly. It is critical for people in our business to understand how and why this happens. And when markets deliver unacceptable outcomes, we turn to the law, which deals with rules developed to do things that markets cannot. As the Internet has changed from a research tool to a public utility over the past 25 years, legal questions have become increasingly important to computer scientists. We need to understand how we can become liable for things we do online, and make only the promises we intend to. We need to understand online harms – from scams through hate speech to monopolies – and the regulatory responses that are likely or even possible. Finally, ethics is not merely a requirement for the professional accreditation of our degree course, but increasingly important for working engineers in a world where laws often lag fifteen years behind technology.
In terms of syllabus structure, this course prepares you to study security, systems, business and e-commerce in Part 2. Ethics is also a requirement for professional accreditation of your degree, and many of the key ethical questions are best understood in the context of the relevant law and economics.
An overview of each lecture is provided below. If you would like to go beyond the standard material, some suggested readings are provided, although I encourage you to also delve into areas that you find particularly interesting. I have also provided links to prior Tripos questions that relevant to each lecture (please note that some questions relate to multiple topics). The solution notes for the questions are available here.
I would also like to acknowledge Professor Ross Anderson who designed and previously lectured this course.
Lecture 1: Classical economics and consumer theory (3 November 2020)
I will begin by providing an overview of the classical view of economics: that under certain assumptions, markets provide an optimal way of allocating resources. This view had its roots in Adam Smith's Wealth of Nations and was further developed by writers such as Ricardo and Jevons to explain forces driving the industrial revolution. We'll explore concepts such as comparative advantage, marginal utility, opportunity cost and exchange, so that you get an idea of what's 'under the hood'. We will then look at the limits of classical economic models, from externalities to the fact that efficiency, welfare and justice are not always the same thing.
Suggested additional readings. There is a huge literature on basic economics. Cambridge economics students cut their teeth on Varian's textbook, 'Intermediate Microeconomics', of which your college library should have many copies. You might look at chapters 1-6 and 14-16 to begin with. For the perspective of a Keynesian economist who advised President Kennedy, try JK Galbraith's 'History of Economics'; for a more modern text that covers technology and sustainability as well as the classical basics, see the core econ textbook (especially chapters 8 and 11).
With the Brexit vote, the Trump election and the growth of anti-globalisation protests, trade has become a controversial topic; the forced liberalisation of India's trade in 1991 provides some useful data on the value of free trade. Another topical issue is the Great Recession since the crash of 2008. The best survey of the effects of credit crunches historically has been written by Reinhart and Rogoff. The Great Recession has also been blamed on globalisation, and on the Internet disrupting one business sector after another; Andrew Odlyzko studies technology business cycles and discusses interesting parallels between recent events and the railway manias of early Victorian England.
Of enduring interest to computer scientists are some more modern criticisms of the classical approach that fall within the realm of microeconomics rather than macroeconomics.
Prior tripos questions: 2020p7q3; 2007p3q1
Lecture 2: Information economics (10 November 2020)
Information goods and services markets tend to be characterised by high fixed costs, low marginal costs and increasing returns to scale, together with lock-in effects, all of which tend to lead to monopoly or oligopoly. In many markets, there are also network effects: the more people use a given service, the more value it is to each user. While products may take a long time to reach critical mass, once they do, they take off very rapidly (as happened with faxes in 1985-88, email in the late 1990s and social media in the late 2000s). Network effects can reinforce a tendency to monopoly.
The strategies used by monopolies to maximise their revenue are important, both as a practical foundation for later work on e-commerce and as a theoretical underpinning for understanding regulation (and the antitrust cases that successful tech companies often attract). Monopoly strategies include market segmentation, price differentiation and bundling. The social and other costs of increasing levels of monopoly are becoming ever clearer, while the possible public policy responses are a work in progress, and the leading edge of the debate may be here or here. See this special section in The Economist for a broad up-to-date overview.
Suggested additional readings. The classic reference on all this is Shapiro and Varian's 'Information Rules'. Varian's textbook also has some useful material, especially in chapters 32-36. As for online resources, there are many, from Wikipedia through the 2014 Nobel lecture by Jean Tirole. The most relevant chapters of Core Econ are 16 and 21, and there's an up-to-date survey in Joseph Stiglitz' The Revolution of Information Economics: The Past and the Future. Andrew Odlyzko has an interesting paper on price discrimination and the erosion of privacy.
Prior tripos questions: 2016p4q7; 2009p3q5; 2008p6q1
Lecture 3: Market failure and behavioural economics (12 November 2020)
We will look at a number of other ways in which information goods and services markets can deviate from the classical ideal. These include asymmetric information, where one party to a contract knows more than the other. For example, people applying for health insurance typically know more about their health than the insurance company does, which leads to adverse selection effects: sick people buy more cover. (Attitudes to risk in general are highly perverse; see for example John Adams on Cars, Cholera and Cows.)
Clickbait feeds on human irrationally. The public misperception of risk is a particular problem: people worry too much about terrorism, for example, and about child safety, while less personal hazards from cyber-crime to global warming get discounted. There is now a thriving field of 'behavioural economics' at the boundary between economics and psychology that seeks to explain systematically irrational behaviour in terms of the perceptions and biases that we acquired in the course of our evolutionary history. Policymakers have learned in the past decade to make choices easy, attractive, social and timely — just as firms try to make the easy choice the most profitable one.
Suggested additional readings. Advertisers have used behavioural techniques for years, and they shade over to hustlers, fraudsters and demagogues. Our innate biases are modulated by culture, and can be transmitted or even amplified by carelessly designed systems. Here is a paper by Calıskan showing how machine translation systems inhale sexism, racism and homophobia along with their training data. There's also a nice talk, Phishing for Phools, on the abuses of behavioural techniques in marketing by the Nobel prizewinner George Akerlof; the most thorough introduction may be Thinking, Fast and Slow by Danny Kahneman while the history of the subject is in Misbehaving by Richard Thaler.
Prior tripos questions: 2020p7q2; 2018p7q2; 2012p4q7; 2007p3q1
Lecture 4: Auction theory and Game theory (17 November 2020)
Auctions drive much of the Internet, from the ad auctions that power Google and Facebook to the more obvious offerings such as eBay. They have been around since at least the times of ancient Greece, and are the traditional way of selling art, livestock and much else (for the Dutch flower auctions, see here). Much of the dotcom boom was based on the premise that online auctions would become the dominant means of doing business in many sectors. eBay duly grew from nothing to blue-chip status in a few years; Google took over the ad market; and the UK government made billions from auctioning off spectrum for third-generation mobile phones. Spectrum auctions have continued worldwide, becoming ever more complex; they are an example of combinatorial auctions, a problem that is not only NP-complete but applicable in many areas of practical interest.
A surprising number of things can go wrong. The British government's success was not replicated everywhere else; in a number of countries, phone companies managed to rig the auctions and get bandwidth cheaply. Often this didn't require any overt criminal behaviour; the rules of the auctions were such that players could signal to each other, during the bidding process, which blocks they were interested in. The resulting tacit collusion meant that the taxpayers in many places got much less than expected. (See Paul Klemperer's 'What Really Matters in Auction Design' for the practicalities, his 'Guide to the Literature' for the Revenue Equivalence Theorem, and 'Why Every Economist Should Learn Some Auction Theory' for its applications.)
It is important to understand how ad auctions work, and the side-effects they can have. Ads are sold by second-price auctions where the bid price (the cost per click) is multiplied by the ad quality (click-through rate, relevance) so that the ad platform can maximise its revenue. Here's a simple description from Bing and a more rigorous treatment from Google's Chief Economist (with a video). Now weighting bids for search ads by quality may seem innocuous, but with social media this approach can end up weighting ad bids by virality. A former Facebook insider reported, for example, that Trump paid less for ads than Clinton in the 2016 election campaign as his ads were more engaging as clickbait. More generally, ad auction mechanisms that favour clickbait can drive polarisation by rewarding extremes in ways that troll farmers can monetise.
Game theory deals with such fundamental issues as whether people cooperate or fight to achieve their goals. One of the classic puzzles in game theory is the Prisoner's dilemma. Two crooks are arrested and questioned separately about a robbery. The police tell each of them that if he confesses, he will go free while his partner will get 10 years for the robbery. If he keeps quiet and his partner confesses, it will be the other way around. If both confess, they will get five years; while if neither confesses, they will get a year each for possessing a firearm. Here, the optimal strategy from the prisoners' collective viewpoint is for both to keep quiet, but if they cannot both trust the other then the optimal strategy for each individual is to confess.
This is typical of many problems encountered in real life. Resolving them is easier where the games are not isolated, but are part of a series. Then one might, for example, have a strategy of tit for tat – if you cooperate with me this round, I'll return the favour next time; but if you stab me in the back, I'll retaliate. Being able to reason about strategies in a quantitative way is important not just in economics, but also in fields such as evolutionary biology where game theory provides elegant models of much social behaviour in animals.
At the level of routine economic analysis, we get useful tools for understanding monopoly and oligopoly behaviour. For example, suppose that it costs $250 to fly a passenger from Boston to London, and only two airlines compete on the route. How will they set prices? Will they collude and charge $500 each, making a healthy profit, or will they compete for market share and charge $300 or even $255? What sort of strategies are available, and what sort of equilibria might emerge? Given the small number of firms in the typical IT market sector, such issues matter.
At this stage, I suggest you read Ross Anderson’s Economics chapter, from the third edition of his Security Engineering book. Ross Anderson developed much of the course material for Economics, Law and Ethics, and this chapter provides a good overview of the lectures covered so far, including auctions and game theory.
Suggested additional readings. There are many relevant books and web pages, such as the Stanford Encyclopaedia of Philosophy and chapter 4 of Core Econ. And with the Brexit negotiations showing the difficulty of reaching a deal between two sides that are themselves divided, see the theory of two-level games which models just that. Of particular interest to computer scientists is the work on the evolution of cooperation by Bob Axelrod, who initiated regular tournaments of iterated prisoner's dilemma that also led to the development of genetic programming ideas. Finally, we also see game theory in game shows; there's some fun here.
Prior tripos questions: 2019p7q2; 2011p4q7; 2010p4q8
Lecture 5: Principles of law (19 November 2020)
There will be two talks on legal topics. The first, will cover the basics. As the syllabus puts it, these are: criminal and civil law, contract law, choice of law and jurisdiction, arbitration, tort, negligence, defamation, and intellectual property. The gloss on that is: what do you have to do online in order to incur liability, or to impose it on someone else; and where can you be pursued, or pursue them, through the courts once you have done so?
Intellectual property is sometimes touted as the foundation of prosperity in the information age, but has long been controversial. Powerful lobby groups, such as Hollywood and the music industry, have pushed for increased legal protection in ways that have brought them into conflict with the computer industry and with digital-rights groups. And some of the lab's most successful startups, from RealVNC to Xensource, have made versions of their products available as free software. Free products such as Linux, Apache and gcc are shared infrastructure on which everybody builds and to which many contribute. Meanwhile, as critical components of most services move to the cloud, firms are less dependent on controlling the rights in their own freestanding software components.
Suggested additional readings. There are often fascinating IP issues raised by technological progress, and they can derail major projects; see for an example the case of Google Books. The music and film industries have lobbied hard for tougher copyright laws but it's not clear that that's done them any good. On the other hand, as we get software in everything, classical property rights are being eroded. In 2018, the US Copyright Office changed the rules to allow people to break DRM to fix stuff (though selling tools to do that remains illegal). And while some tech firms invest heavily in patents, Microsoft has just open-sourced its entire patient portfolio. So if you're planning on starting a business, you may want to think hard about your IP strategy.
Prior tripos questions: 2016p4q7; 2006p4q1
Lecture 6: Law and the Internet (24 November 2020)
The second talk relating to law, by Richard Clayton, looks at more technology-specific aspects of law and regulation. A number of EU directives affect how you can do business on the net, covering subjects that range from distance selling, electronic commerce, data protection and electronic signatures to copyright; most will persist after Brexit, as UK firms won't otherwise be allowed to sell in Europe. There are also specific UK laws, such as the Investigatory Powers Act, that you might have to watch out for.
Prior tripos questions: 2020p7q2c; 2019p7q3; 2017p4q7; 2014p4q7; 2013p4q7
Lecture 7: Philosophies of Ethics (26 November 2020)
Technology is moving so quickly that the law usually lumbers along ten or fifteen years behind. Lawmakers are not always the most geeky members of society, so the laws they make often don't fit that well with the tech we build. So laws alone cannot provide a comprehensive guide for action, except possibly for the rapacious. Ethics and social norms take up some of the slack. Many ethical debates have live connections to economic arguments, and some of them will no doubt crystallise into laws in due course. The Nuffield report mentioned in the lecture explores this in the context of medical privacy and research ethics; it may be found here.
Suggested additional readings. Much has been written on ethics in the last few thousand years, yet interesting new debates are opening up in topics from neuroethics to policy. Behavioural psychologists are exploring the evolutionary roots of morality; here is a video by Jonathan Haidt on the topic, suggesting an evolutionary basis for liberal and conservative mindsets. Finally, if your third-year project involves collecting data from human subjects, you had better think about how you do an ethical assessment and get approval for it if necessary.
Prior tripos questions: 2017p4q7; 2015p4q7
Lecture 8: Contemporary ethical issues (1 December 2020)
This final lecture covers contemporary ethical issues, selected due to their relevance to computer science and the development of new technologies. These issues include algorithmic decision making and bias, including predictive policing; environmental impacts, particularly relating to end-of-life and e-waste; the power of private platforms; and privacy, including health privacy. The lecture ends by discussing ethics in the workplace, specifically what employees are doing to ensure their work does not cause harm, and how companies are responding.
Suggested additional readings. Liberty have a report on the use of predictive policing in the UK. Mascarenhas et al. provide a useful overview of the ethics of e-waste management. If you would like an insight into the Cambridge Analytica scandal, then I suggest you read Christopher Wylie’s book Mindf*ck, which engages in issues relating to the power of private platforms, such as gamification and engagement and the inherent information asymmetry. A broad view on the ethics of big data research is here and a critique of unethical social media practices can be found here. Another subject is the attention economy; although this phrase was coined by Herb Simon, there's a lot of recent work, including that by Tim Wu. A further new development is that we can now do experimental ethics at scale; and Internet adoption appears correlated with an increase in political partisanship. Yet another is the digital divide; a few years ago, people were concerned that the poor did not get adequate access to the Internet, while now the concern is that, with firms battling to make their apps ever more addictive, poor kids spend too much time online.
Prior tripos questions: 2014p4q7
Problems for supervisions
For each supervision I recommend that you select at random two past exam questions on the material already covered: exam questions since 2003 can be found here. If you want smaller, more mathematical questions of the type familiar from other courses, see the revision questions in Varian's textbook, chapters 1-6, 14-17, 24-25, 27-28 and 32-36, and the problems in its companion volume `Workouts in Intermediate Microeconomics'.
However, one word of warning: many Part 1B students may never have studied a humanities subject since GCSE. It is a different task from learning a programming language; it is not sufficient to acquire proficiency at a small core of manipulative techniques, and figure out the rest when needed. Breadth matters. You should spend at least half of your study time for this course on general reading. There are many introductory texts on economics and on law; your college library is probably a good place to start. And it is really important to write essays, to get practice at marshalling your arguments and presenting them coherently. Give your supervisor essays to read and ask him or her to give you feedback.