Digital Ethics FAQ
- What is digital ethics?
- What is an automated decision-making system?
- Who uses automated decision-making systems, and what are they used for?
- What is an algorithm?
- What’s the relationship between automated decision-making systems, artificial intelligence, and machine learning?
- What is Explainable AI/XAI?
- What do ethicists do?
- Why is digital ethics important?
- How can automated decision-making systems be used for good?
- What is the goal of the Jain Family Institute’s digital ethics and governance initiative?
- What is the goal of the Jain Family Institute’s digital ethics and governance initiative?
- Who is the audience for the Jain Family Institute’s research in digital ethics?
- What is the Jain Family Institute doing to help develop digital ethics education?
- What is the Jain Family Institute doing to help develop digital ethics governance and policy?
- What other organizations are working on similar issues in digital ethics?
What is digital ethics?
Digital ethics is a field that applies ethical reasoning to questions arising from digital technology. This broad term encompasses algorithmic ethics, ethics of AI, data ethics, ethics of machine learning and data science, ethics of technology, ethics of systems engineering, and more. It’s useful to keep in mind that this is a new field: terms are mutable, new questions arise every day, channels of communication are still being built between computer scientists, ethicists, legal scholars, and so on. JFI’s work in digital ethics focuses on automated decision-making systems.
What is an automated decision-making system?
An automated decision-making system (ADS) is any procedure that automates or computerizes some of an organization’s decision-making in novel ways, typically by either (a) replacing steps that involve discretion or judgment, or (b) implementing new inference and decision procedures that do not have counterparts in existing processes.
Who uses automated decision-making systems, and what are they used for?
Automated decision-making systems (ADSs) are used by all kinds of organizations, including governments, social media platforms, insurance companies, businesses, and development organizations. For example, ADSs have been used to determine eligibility for welfare benefits, decide whether to grant bail to a defendant in criminal court, triage incoming emergency calls, prioritize building inspection, identify instances of tenant harassment, improve rescue operations and disaster management, monitor food insecurity, detect pneumonia in radiographs, assign students to public schools, and assess teacher performance.
What is an algorithm?
An algorithm is a description of a procedure that a computer can follow (at least in principle). There are many different families of models, or kinds of algorithms, that data scientists and programmers use to approach questions of prediction, estimation, matchmaking, and classification. Rule-based algorithms, for instance, are straightforward implementations of the kinds of rules that human decision makers follow. Machine-learning systems use procedures that are themselves selected by computers to fit the problem at hand; these can involve millions of steps that a human would find strange, but which are sometimes even better than humans at finding patterns and making predictions.
What’s the relationship between automated decision-making systems, artificial intelligence, and machine learning?
Artificial intelligence is a broad term that refers to systems that are capable of performing tasks that we ordinarily think of as requiring human judgement, perception, or intelligence. Some applications include voice recognition, self-driving vehicles, industrial process monitoring and control, interactive computer assistants, predictive typing, and games.
An automated decision-making system (ADS) is a use of artificial intelligence in the context of an organization’s decision making process, often in ways that supplement or replace human decision makers. So, for example, a very simple ADS might use a series of explicit rules that decide whether online credit card applications should be approved, denied, or sent to a reviewer.
Machine learning refers to algorithms, types of algorithms, strategies for creating algorithms, and systems that run on algorithms that are much more sophisticated than simple rule-based systems. These algorithms usually “learn” in the sense that they use data to figure out what rules to use. For example, an ADS that involves machine learning can be trained on data, like past parole decisions, and make decisions about circumstances that have not been explicitly programmed in the system, like parole judgements about future candidates.
What is Explainable AI/XAI?
One of the key issues facing automated decision-making systems (ADSs) is that decisions are often made within a ‘black box’: If an ADS makes a decision, it can be almost impossible to know what factors played a role in that decision, and to what extent those factors mattered. Explainable AI or XAI refers to techniques that attempt to make the decision procedure involved in an ADS verdict more intelligible to those affected by the systems.
What do ethicists do?
In general, ethicists are interested in what people ought to do. They are not primarily interested in what people think they should do or what people actually do. From their perspective, these are issues for sociology, psychology, and social science. Instead, ethicists try to determine what people should in fact do. So, for example, ethicists don’t ask, “Do people think they should lie to make someone feel better?” or “Do people lie to make others feel better?” — these are questions of fact. Instead, they ask “Should I lie to make someone feel better? Should I tell the truth about my credentials on my resume? Should I refrain from driving gas-guzzling cars?” These are questions of ethics — and questions like these are at the forefront of an ethicist’s mind.
Some ethicists are specialized: bioethicists try to determine what we should do in medical contexts; environmental ethicists study what we should with respect to the natural world; and business ethicists develop frameworks for good practices in commerce. Digital ethicists study how we should develop, use, and limit new technologies like ADSs, and what the implications — good and bad — are of using these systems in particular contexts.
Why is digital ethics important?
As automated decision-making systems (ADSs) are increasingly deployed alongside or in place of human-operated processes, questions of how to ethically create and utilize them arise. Digital ethics aims to address the lack of research on the societal impacts of these technologies, and to begin the process of updating regulation and governance standards to accommodate these new domains.
The questions are ethical, technical, and political. Are there applications for which it is unethical to use ADSs? What sorts of tradeoffs and error tolerances are we willing to accept for different problems? What determines whether a decision is fair, and how can we ensure that ADSs are used in ways that will lead to fair results? Digital ethics is a new, interdisciplinary field which aims to answer these questions.
How can automated decision-making systems be used for good?
Automated decision-making systems (ADSs) have the potential to improve decision-making on the part of governments and firms in a number of ways. In simple cases, ADSs can use powerful statistical methods to increase the accuracy of agency decisions. More complex ADSs can enable an agency to better harness the available information in order to reach conclusions that the agency cannot reach on its own. So, for example, some cities have used ADSs to flag patterns in tenant complaints to more accurately identify illegal tenant harassment. In addition, by offloading some tasks to computers, governments and NGOs can more efficiently use scarce resources, freeing up public servants to do other work, respond more quickly and effectively to pressing issues, and focus on areas in which humans are indispensable. Moreover, in many cases automated systems may limit the effects of human bias, and offer increased explicitness about decision processes and rationales.
What are the ethical issues raised by automated decision-making systems?
In brief, automated decision-making systems (ADSs) raise ethical concerns around governmental transparency, nondiscrimination, due process, and data privacy. Jain Family Institute fellows explore the following questions, among others:
Explainability: What counts as an explanation of an ADS verdict? Will a single kind of explanation suffice for our use of ADSs across specific domains, like legal proceedings, credit scoring, or university admissions?
Governance: What kinds of institutional structures should be in place for the administration and oversight of ADSs?
Fairness: To what extent, and how, should considerations about bias, discrimination, and fairness factor into ADS design?
Autonomy and Manipulation: What provisions should be in place to protect citizens from manipulation by ADSs? Are there distinctive concerns about manipulation that must be addressed before more widespread adoption of ADS?
What is the goal of the Jain Family Institute’s digital ethics and governance initiative?
JFI aims to open the debate around ethics, governance and standards for automated decision-making systems to the broadest possible array of stakeholders. This way, communities can make informed judgements about the appropriate use of such systems in government, business, and media practices.
The Jain Family Institute believes that addressing the ethical implications of the widespread use of automated decision-making systems urgently demands the development of novel interdisciplinary expertise.
Who is the audience for the Jain Family Institute’s research in digital ethics?
JFI’s work responds to the urgent need for accessible literature on automated decision-making systems so that the public—including journalists, activists, lawyers, and civil servants—can identify problems and explore solutions on their own. Our materials are not be focused on pursuing a fixed agenda or set of solutions. Rather, they leverage JFI’s unique combination of capacities in computer and data science, philosophical ethics, public policy, history, the social sciences, pedagogy, and data visualization to develop research that opens the discussion to citizens that are not currently part of the debate.
What is the Jain Family Institute doing to help develop digital ethics education?
JFI has encouraged the development of novel pedagogical approaches in ethics education, most notably through support for our fellows affiliated with the ‘Integrating Ethics Education into New Engineering Education Transformation’ (NEET) program at MIT and the Embedded EthiCS program at Harvard University. You can read more about some of our fellows’ work here and here. We have also supported the development of public education materials, like this website developed by students at Harvard’s Kennedy School of Government in consultation with JFI staff.
What is the Jain Family Institute doing to help develop digital ethics governance and policy?
JFI has provided research support for municipal governments who seek to develop frameworks for regulating and encouraging the responsible use of automated decision-making systems. We are working with non-governmental development organizations to better understand the challenges facing those who aim to use AI for good and look forward to bringing those insights into policy debates about this technology.
What other organizations are working on similar issues in digital ethics?
Other digital ethics research organizations include Data and Society, The Future Society, MIT Media Lab, Berkman Klein Center at Harvard, Future of Humanity Institute at Oxford, Leverhulme Center for the Future of Intelligence at Cambridge, and AI Now Institute.