Software for Tyrants

PolicyMaker’s gimmick — what the Harvard marketers call its “algorithm” — is devilishly simple. You just punch in your preferred policy outcome, a laundry list of affected “players,” and a potpourri of potential “obstacles,” and seconds later your PC disgorges page upon page of brightly colored “feasibility graphs” showing the pros and cons of dozens of diverging, pre-programmed political strategies for achieving your goal.

Since the debut of the prototype two years ago (version 2.2 just hit the stands), PolicyMaker’s influence on international policy-making has been dizzying. It has been used to set local health priorities in Tanzania, to analyze government tobacco policy in Vietnam, to assess national pharmaceutical policies in nine African nations, and to impose major health sector reform in Zambia, Mexico, and the Dominican Republic. Professor Reich has even heard a rumor that Zambia’s corruption-riddled government is using the program for “political election purposes.”

Professor Reich, who teaches international health policy, concedes that PolicyMaker, and computer-assisted political analysis in general, carry certain risks. “PolicyMaker is designed to help policy makers get what they want, which is not necessarily good from an ethical or a technical perspective,” Reich says in a statement that accompanies the program. Coming up with something you can sell to the public “should not substitute for assuring the ethical and technical bases of a policy.” And, of course, even totalitarian regimes can benefit from the software. As Reich puts it, “It can be adapted to situations with different distributions of power.”

PolicyMaker is just the tip of the iceberg when it comes to the brave new world of techno-politics. Step inside any school of public policy today and you will find that policy-making is the province of automatons. Focus groups, online polls, computer programs galore — public policy education boils down to process.

New software like PolicyMaker is helping drive the revolution. Microsoft Project, for example, is a popular generic policy planning tool. There are programs for “stakeholder analysis” and “political mapping” and infrastructure design. The United Nations Development Program has its own software; and Private Agencies Collaborating Together (PACT) has developed, with USAID funding, an “analytic capacity assessment tool” for non-governmental organizations called DOSA (Discussion-Oriented Self-Assessment).

All these initiatives pale in comparison with what lies ahead. Already under development is software that allows users to assess the prospects for generating political will — out of thin air! — for a virtually limitless range of domestic policy reforms. Experts in the field of artificial intelligence (AI) anticipate a new generation of computers that will soon be able to act as autonomous agents, to assess the political environment and even think using emotions and instinct. Prototypes have already been made that purport to resolve disputes in bioethics and other fields, grade academic essays, and manage prison sentencing. These new tools are worlds apart from the high-speed calculators of yore; for they are programmed to make inherently political value judgments — to prefer, say, contracting out over busting unions. What’s more, these new systems, if and when they fail, cannot be taken to task in public hearings or congressional inquiries.

All of which signals a major step beyond the last generation of artificial intelligence, in which computers merely provided conclusions, albeit powerful ones, based on the information human beings fed them. Examples of these earlier information-processing tools include medical systems such as MYCIN, which tabulates information about patients’ cases (e.g. history, symptoms, laboratory test results) to help physicians formulate hypotheses and nurse managers control costs; military systems such as the Integrated Defense Systems (IDS), to determine optimal response times and simulate enemy counter-measures; criminal law enforcement software such as Brainmaker, used to identify police officers who display behavior that could lead to crime or corruption; and psychiatric tools such as the somewhat scary-sounding Good Mood Program, which enables users to talk through their personal difficulties.

For those who believe that efficient delivery of public services is the most important goal of public administration, these twenty-first-century tools have clear advantages. First, they remove inconvenient levels of discretion from the policy-making process. That might appeal to liberal political theorists such as Theodore Lowi (The End of Liberalism, 1969) who have argued passionately that the American system delegates too much discretionary authority to administrative agencies, a recipe for patronage government and unaccountable bureaucracies. And liberal idealists who object to a world where decision-makers are guided primarily by self-interest doubtless welcome computer-assisted policy-making. Finally, those who see public administration as a branch of science must feel at home in the new world.

But to many, the rise of computerized policy-making is troublesome. It throws to the winds the Aristotelian directive to discover the common good through careful deliberation and honorable leadership. It undermines the civil servant’s power to challenge elected officials — by raising questions, courting public approbation, or even taking some initiative to bring about a change of course (what political scientist Hugh Heclo has called “loyalty that argues back”). And, more generally, it erodes the role of public servants as individuals — as thinking people and ethical agents. For all its snazzy graphics, mechanized policy-making puts us on a dangerous path.

Neil Seeman writes editorials and essays for the National Post, a Canadian daily.