Watson should have run for US president: How smart algorithms are better than dumb politicians

Quartz | January 24, 2017

The best candidate for the office of president of the United States wasn’t on the ballot last November. It was Watson, IBM’s cognitive computing engine. Watson would be a president of intelligence and integrity, temperament and dedication, working round the clock to devise solutions to our toughest domestic and international challenges. In stark contrast to any human leader, Watson is a not a political animal or stylish personality, rather he is a policy machine made of pure substance. For a brief flicker in early 2016, a private foundation not affiliated with IBM actually launched an online campaign advocating “Watson for President”—too bad it wasn’t taken seriously. America’s political system needs much less Hillary Clinton and Donald Trump and much more Watson.

Western democracies are succumbing to populist movements promising little more than nostalgia. Jobs are supposed to “come back” from abroad that were actually lost to technology, or won’t exist within a few years. The paradox of populism is that it makes narrow appeals to aggrieved segments of the population—appeals that make little sense on a national scale. It glorifies the interests of a minority over the majority.

The political class, rather than combating these tendencies, has succumbed fully to them. Hillary Clinton’s flip-flops on the Keystone XL Pipeline, Trans-Pacific Partnership, and immigration are just a few examples of how even the more qualified presidential candidate—who surely knew better—changed tune in order to please specific voter demographics rather than sticking to what is in the collective national interest. And still she lost.

Since there is no pillar of society left defending the honor of reason, I propose we use technological automation to get rid of some politicians’ jobs as well. We don’t need most of them. We need facts, analysis, scenarios, and strategies. We need to consult the public, study proposals, model outcomes, weigh costs and benefits, factor in potential reactions and consequences, and make long-term decisions. This sounds like a job for experts, not hacks.

This is where Watson comes in. IBM began to show off its AI prowess when its Deep Blue computer took on Gary Kasparov in a series of chess matches two decades ago, scoring its first win in 1997. By 2016, IBM’s cognitive machine had defeated the world’s leading Go champion, China’s most complex board game. Watson’s analytical tools are currently deployed in hospitals personalizing cancer treatments, in school systems tailoring student curricula, and in China suggesting smog reduction strategies.

Watson was also watching the 2016 presidential debates, using its tonal analysis tool to compare them with videos going back to the Kennedy-Nixon 1960 debates, and found that the level of civility amongst candidates has declined markedly over time. At the same time, IBM’s Debating Technologies searches massive troves of research to generate and concisely summarize the pro and con positions on thorny issues such as pipeline projects, tax policy, and the regulation of violent video games. Rather than fact-free debates, bringing data into deliberation makes democracy more rational, rather than overly emotive.

Watson can also be a constructive tool for shaping major national policy such as infrastructure, healthcare, energy and drugs, or calculating the costs and benefits of various policies while studying cases drawn from all over the world. Watson’s research advocates proposals—such as single-payer healthcare, free university education, legalizing recreational drug use, upgrading infrastructure and public transportation, raising the minimum wage, and investing in renewable energy. By analyzing data and providing insight, Watson doesn’t replace government, but makes it smarter. IBM vice president Guruduth Banavar recently told CNN that, “Cognitive computing is about partnering between machines and humans, combining their strengths to solve big problems.”

Now that we have technologies that help us think, we need a government that’s designed to act.

As things stand, however, America’s government has too many representatives and too few administrators, too many lawyers who debate and too few professionals who actually do things. A better American system would be substantially more technocratic without being any less democratic. In my new book Technocracy in America, I argue that the ideal government for the information age is a “direct technocracy,” combining the direct democracy of Switzerland—where citizens dictate the national agenda through frequent plebiscites and initiatives—with the technocracy of Singapore, whose impartial civil service constantly researches how to improve the effectiveness of policies that benefit the public. As radically different as these two small countries are, they are the only two nations in the world that rank in the top tier of many important indices such as the WEF’s Global Competitiveness Index, Infrastructure Quality Index, and Sustained Prosperity Index; INSEAD’s Global Innovation Index; and the World Bank’s Government Effectiveness Index. They have higher median incomes and life expectancy, and less corruption and unemployment than America.

Both countries also use a combination of data and democracy to steer governance far more effectively than America does. Census data, surveys, petitions, public consultations, and many other tools play a role in tweaking policies around taxes, immigration, education, retirement, and more. Rather than fact-free debates about stale ideological dichotomies such as “big” versus “small” government, they think issue-by-issue about how government can be most useful. In the US, by contrast, data-gathering has been much more a tool of politics than policy. The Democrats in 2012 and Republicans in 2016 used geo-located predictive polling to cleverly target campaign messages in battleground states. But in big long-term challenges like raising the savings rate, reforming education curricula, or mapping out national infrastructure, America is barely out of the starting gates.

The Obama White House had made some efforts to beef-up the role of data and research. In 2016, the Office of Management and Budget (OMB) created the Commission on Evidence Based Policymaking to promote better data gathering and statistical analysis in shaping regulation. America’s former CTO Todd Park and his successor Megan Smith were also strong advocates of integrating data tools into government agencies both to streamline them and encourage data-sharing across them. They also oversee a new Digital Service Corps of young tech talent on sabbatical from the IT industry who are the boots on the ground of integrating new technologies into old bureaucracies. But even as they have tried to make federal services such as Obamacare and veterans affairs more efficient, they have faced enormous bureaucratic barriers, multiple data systems across agencies, and conflicts across federal and state lines such that national e-government remains a distant dream.

In a good technocratic system, the civil service are the honest stewards of national data rather than consultants, contractors and pollsters. Other countries have engineers in politics; America has barely a handful. One explanation offered by science historian Edward Tenner is that distrust of government and a strong private sector have lured engineers away from government. Yet government needs more engineers—fixers and plumbers—rather than financiers. Google’s Hal Varian once claimed that statistician is the sexiest career of the 21st century. In Singapore, at least, he is correct.

Good governance often comes down to a combination of statistics and logistics: Analyzing data and getting things done. Rather than our obsession with political personalities and their net worth, we should make sure that whoever is elected or appointed to senior positions is trained in the technocratic art and science of using data to enhance governance for the people.

One step in the right direction would be more data availability on the workings of government itself. Civic initiatives such as GovTrack and Project Vote Smart allow anyone to go online and view in real-time all legislative sessions, bills up for authorization, statements made in hearings, voting records of Congressmen and, crucially, all available financial records of their campaign contributions. Such public-disclosure campaigns strengthen the hands of citizens to ensure their priorities are addressed. When survey data reveals that the elderly require greater choice in medical programs, and youth more education options, the action items should be more than talking points for the next campaign. A more digital society advances transparency in the service of accountability, without which democracy is just chatter.

There could hardly be a more pro-democratic technocratic measure than mandatory voting. Rather than politicians getting elected with 30% of the total voting-age population’s approval, mandatory voting guarantees a demographically inclusive process, and eventually a more informed populace as well. The US could save itself billions of dollars in voter registration drives, mechanical polling booths and other Election Day drama if American citizens at home and abroad had a one-week window to cast their votes electronically through a secure online portal. Estonia has had national internet voting since 2005, while e-voting is currently being rolled out across Switzerland. Leapfrogging to direct digital democracy is a simple matter of passing a mandatory voting law plus building an app.

But voting alone is far from the best means of capturing popular sentiment on an ongoing basis about the vast range of issues that concern citizens. For that we need more qualitative data from surveys and social media, and quantitative data such as demographic and economic trends. Data can be more comprehensive than election results for it is broader in scope (covering the full spectrum of issues rather than being hijacked by hot-button topics) and fresher (collected more regularly than infrequent elections). Scaling technology is easier than scaling trust, but the former can be a path to the latter. More substantive interactions—even virtual ones—between citizens and politicians could make a massive contribution toward reducing the trust gap in American politics.

Today we think of data tools as aiding democracy, but eventually, democratic deliberation (whether elections, initiatives, surveys, or social media) become contributing data-sets among many that together help the government steer policy. For example, data that represents the unrepresented (those who don’t actively vote or participate in surveys)—such as their financial behavior and education status—are essential inputs for leaders to ensure they are taking everyone’s needs into account.

Data-driven direct technocracy is superior to representative democracy because it dynamically captures the specific needs of the people while short-circuiting the distortions of elected representatives, special interests and corrupt middlemen. In Switzerland, democracy is not something that is done for the people by representatives; they co-create and co-design policies. This means that the frequent Swiss initiatives and referenda are not lesser events in between more important elections; they are a form of habitual voting on all issues of importance.

Both the Swiss and Singaporean examples prove that we don’t have to be ruled by data, but can balance it with democracy so they complement each other. Data can determine which policies are necessary, while democracy can modify and ratify them.

At a time when Americans have lost their shared view of reality, this is the juncture at which America’s political system needs a managed evolution. Many commentators have remarked that this election was the first time they’ve felt that America has become two countries, two divided voting blocks that simply hate each other and talk past each other. The electoral map has turned America into 435 battlegrounds rather than one United States. Listening to the campaign, one could mistaken for thinking that the American electorate was 300 million factory workers in the Rust Belt rather than the overwhelmingly services oriented economy the country actually is. Less than 10% of American workforce is employed in manufacturing while most of America’s 80 million millennials subsist in the gigonomy of multiple simultaneous jobs they find out about on their smart phones.

It is precisely at this moment then that we need big data to remind us that we are a big country and need a national conversation about what policies will bring the greatest good to the greatest number, and what states can learn from each other. Take trade. No country has reaped more financial gains from global trade than the United States, both in terms of cost savings from importing cheaper goods as well as profits from selling goods abroad. Furthermore, even in Ohio, jobs created by foreign investors pay higher wages than American companies do and provide better skills as well. And yet Donald Trump opposes NAFTA and the Trans-Pacific Partnership that would have forced Asian countries to reduce their subsidies and allow more American firms to compete. In the hands of politicians, the fact that more manufacturing jobs are lost to technological automation than trade is conveniently overlooked in favor of equating trade deals with job losses. Scapegoating foreigners is a better recipe for winning elections than admitting that Washington hasn’t provided workers with the skill retraining programs they were promised four presidential elections ago.

Most of all, then, we need Watson to get ourselves educated. Americans are already geographically and historically illiterate. The media, for its part, has only compounded populist cleavages. The combination of electoral gerrymandering and soundbyte politics sowed the seeds for “filter bubbles” long before Twitter and Facebook came along. Indeed, “real media” is far more to blame for a divided society than “fake news.” It’s not Facebook or Google’s fault that the news media executives gave far more attention to Hillary Clinton’s private e-mail server than to any policy issue. Let’s also not forget that it’s big media companies who have set-up white label content creation companies to help advertisers make their propaganda look like real news in the first place. No wonder a recent Stanford study found that 80% of young people can’t distinguish between sponsored content and journalistic stories–because mainstream media’s side business is booming.

At least Watson can easily tell the difference between real and fake news and won’t indulge in promoting it as our own president does. Unlike lazy humans, Watson would check to see the diversity of sources and references to “news” stories to determine if they were actually corroborated. Rather than blaming Facebook for spreading “fake news,” a Watson oriented society would be anonymously “listening” and analyzing thousands of legitimate discussions on genuinely participatory Facebook pages and groups to better understand public sentiments. China already does this with WeChat and Weibo. Though we shouldn’t condone censorship or surveillance, opinions that we share publicly can reveal what certain demographics are thinking and shouldn’t be ignored.

Given the widely cited decline in civic associations in America today—gatherings where, as Tocqueville put it, “people look at something other than themselves”—social media should become ever more a strategic tool for gathering knowledge about citizens’ priorities. After all, should popular sentiments always be filtered through the geographic lens of political constituencies when society has become so physically and digitally mobile? Americans are social creatures—Washington should pay attention to what they say in their new digital communities.

Government for a nation of millennials and their children has to evolve to cope with the technologies of today and tomorrow. For her part, my daughter, who rules an imaginary monarchy called Zarania located on the unclaimed slice of Antarctica, plans to include Watson in her technocratic cabinet. In the real world, Watson may not play more than auxiliary role in the coming years, but it represents a new generation of technologies that can encourage more reasoned discourse through presenting accurate information and realistic scenarios. Using Watson should become a reflex like talking to Siri or Google Home. These are not auto-pilot technologies that dull our sensibilities but rather tools that lower the cost of getting smart. Thomas Jefferson famously remarked that people get the government they deserve. If we use cognitive technologies wisely, we might finally become smart enough to build ourselves a better government.

Link to article