2018 Speed Conference
September 28–29, 2018
2 W Loop Rd
New York, NY 10044
- Website: https://www.dli.tech.cornell.edu/speed
- Abstract Submission: To apply, email an abstract of approximately 250 words.
- Submission Deadline: June 25, 2018
- Notification of Decisions: July 9, 2018
- Position Papers Deadline: September 7, 2018
Algorithmic oversight is one of the fundamental defining policy problems of our age. Two dimensions of the problem are well appreciated. One is scale, as captured by the pithy phrase Big Data: computers work with datasets far beyond human capacity to gather and process. Another is complexity: computers can develop strategies and derive conclusions that are not easily explained to humans. Alone and together, scale and complexity have unsettling implications for privacy, employment, due process, anti-discrimination, and other important social values.
But there is a third and equally important dimension to algorithmic oversight, one that has gone comparatively unnoticed: speed. The characteristic time-scales of human decision-making range from years (for democratic deliberations) to a few hundred milliseconds (for instinctual reactions). But software is capable of making decisions far faster: in a matter of microseconds. This mismatch poses serious technical challenges and even deeper ones for managing and responding to algorithmic actions. When an algorithm acts so much faster than any human can react, familiar forms of oversight become infeasible.
Speed has been a longstanding, hallmark virtue of digital life. Competitors vie to have the fastest operating systems, the fastest search results, the fastest network links, the fastest everything. Digital products and services will speed business processes and everyday life; they will make us faster and more efficient. But is speed always a virtue when delegating decisions and control to AI in areas of individual and social life that may have direct and profound impacts on quality of life and on the ethical and political values guiding individual actions and societal institutions? If the ideal relation of human to algorithm, or human to AI is one of collaboration rather than abdication, how do we reckon with the question of velocity?
Thinkers and makers from a wide variety of fields are grappling with the problem of algorithmic speed and looking for ways to realign high-speed computer systems with the human capacity to understand and direct them appropriately. Robot builders program delays into robots that will interact socially with humans, so that their reaction times are not uncannily quick. Online platforms have complex algorithmic filters that act automatically and immediately on some uploads and flag others for more deliberate human review; so do network anti-intrusion systems. Financial regulators are trying to eliminate unfair market advantages by making some kinds of trading move as fast as algorithmically possible while slowing others down.
These conversations, however, are mostly taking place in silos, in relative isolation from one another. There is not yet a broad understanding that these different fields are all confronting similar issues and can learn from each others’ experience. This unification is well underway when it comes to scale and complexity. But for speed, it has been, well, slower. We think that the time is ripe for an interdisciplinary academic conference to put speed on the agenda.
Chaired by James Grimmelman and Helen Nissenbaum, this 1.5-day conference will explore the challenges of algorithmic governance in an accelerated age. Participants will be drawn from information science; law; computer science and electrical engineering; communications; economics and finance; philosophy; science and technology studies; and other disciplines. They will discuss current, forthcoming, and potential changes in algorithmic speed in their respective domains and techniques for avoiding, detecting, mitigating, and responding to speed-abetted algorithmic mistakes and abuses.
Helen Nissenbaum, Cornell Tech
Call for Papers
We invite proposals to speak at Speed. We anticipate accepting approximately 10-12 proposals, with no preset allocation among topics or disciplines.
Proposals may be based around one of the six research tracks: warfare, financial markets, labor and manufacturing, autonomous vehicles, information security, and content moderation. Or they can combine more than one of the tracks — or develop an entirely different one. We welcome anything on the conference theme.
Proposals may be rooted in any academic discipline or disciplines, including but hardly limited to computer science, engineering, physics, biology, law, information science, design, communications, economics and finance, political science, philosophy, science and technology studies, history, anthropology, and psychology. Or they may come from professional practice in technology, journalism, the military, public policy, medicine, or business. We welcome proposals from anyone with something to contribute to the conversation.
Accepted speakers will have approximately 20-30 minutes to present their ideas and will then engage in substantive discussion with their co-panelists. Speakers will be asked to provide short (~5,000 words) position papers laying out a significant issue in their area and linking it to their presentation. Position papers may be written specifically for the conference, drawn from previous work, or offered as works in progress. Based on the papers we receive, we may with the permission of the authors seek an appropriate academic publication venue for the collection. All accepted speakers will be provided with transportation to and from New York and lodging while here.
To apply, email an abstract of approximately 250 words by 25 June 2018. Applicants will be notified of the decision on their application by 09 July 2018.
Speed is intended to be an inclusive, welcoming, and diverse event. We expect all participants to behave with collegiality. Harassment will not be tolerated.