Future Tense

Driverless Finance

Advanced financial technologies promise major advances—but the risks are just as high.

A person holds a smartphone displaying stock fluctuations in front of a laptop displaying financial information.
Austin Distel/Unsplash

In March 2018, a driverless car struck and killed a pedestrian in Tempe, Arizona. Following the public outcry after that incident, the developers of driverless cars walked back expectations that these vehicles would soon be sold to the public, accepting that they were not yet “safe enough.” Yet while automating driving stirs up concern, few people seem worried about similar pushes to automate financial services. The use of technology in finance is nothing new, but advances in computing power and data usage techniques have ensured that the latest financial technologies, known as “fintech,” are significantly more sophisticated and autonomous. We are now approaching the era of “driverless finance,” and while the dangers are certainly different than those posed by out-of-control cars, fintech is in similarly urgent need of regulation.

To the extent that anyone has raised concerns about these new financial technologies, they have usually focused on the potential for consumer harm (particularly privacy violations and discrimination in the provision of credit). But the technologies used by fintech firms, including marketplace lending and robo-investing platforms, are more complex and move faster than previous generations of financial technologies. Because these technologies can make decisions on behalf of thousands of accounts (or more) at a time, they also have the potential to make both bubbles and panics more likely. That means that these new financial technologies are harder to understand, harder to stop when things go wrong, and more likely to have an outsize impact on the stability of the financial system as a whole. Unfortunately, there has been far too little discussion of how these new technologies could blow up the financial system, taking the broader economy with it. These technologies are exciting, with the potential to make financial services cheaper, more efficient, and more available to underserved populations. However, the risks associated with these technologies are escaping scrutiny because of this excitement, and also because of our collective tendency to underestimate the costs of financial system failures.

We can look at robo-investing to see some of the problems associated with driverless finance. In robo-investing, algorithms automate the selection and rebalancing of investment portfolios. If masses of people start choosing when to make and sell investments based on the same or similar algorithms, then fluctuations in the prices of stocks and bonds are likely to become more magnified, and the markets even more volatile than they are now. The most sophisticated robo-investing algorithms use machine learning, which means that the quality of their decisions depends on the quality of the data used to train the algorithm. Even with good data, machine learning algorithms’ focus on probabilities means that they are prone to ignoring the possibility of infrequent events, no matter how significant those events could be. And if it does become clear that an algorithm is systematically making a mistake, the technology does not yet exist to teach it to stop making that error.

Because of this complexity, a robo-adviser relying on a machine learning algorithm can be very unpredictable. If something unexpected does happen, it could automatically rebalance all of its customers’ portfolios in a way that seems bizarre to humans, and this could be done so quickly that there is no time for human intervention. This could happen at such large scale that it could create problems for asset pricing in general and even destabilize the financial system as a whole.

Other new technologies come with similar downsides. Take the distributed ledger technology that was pioneered by bitcoin but now is being used to process all kinds of transactions, including the trading of smart contracts. A smart contract is a contract where all the terms are recorded in computer code. The parties to the smart contract are recorded on the distributed ledger, and the smart contract is programmed to execute and enforce itself by transferring funds between those parties—all without the need for any human actors or courts to get involved. Again, the complexity of the programming and the speed of execution could be a problem. Imagine that a smart contract withdraws a huge amount of money from a financial institution in erroneous and unanticipated circumstances, causing that institution to default on its contracts with other financial institutions. The ripple effects could be swift and significant, and attempts to undo the initial withdrawal could be complicated if the distributed ledger on which the smart contract is recorded is managed by a large and diffuse group of people (such a ledger could also be independently paralyzed by unanticipated operational vulnerabilities, bringing all transactions that are processed on that ledger to a standstill).

You may think of “fintech” as the province of small startups, and assume that any problems would have limited impact. But these technologies are in fact being contemplated (if not used already) by some of our largest, most established, financial institutions. Fidelity has a robo-adviser called Fidelity Go. JPMorgan has launched a “JPM Coin” for processing payments, and it runs on a distributed ledger. An important industry group called the International Swaps and Derivatives Association is actively exploring the possibility of “smart derivative contracts.” (Derivative contracts helped blow up the financial system in the last financial crisis.) And it’s not just the established financial players—the tech giants are also eyeing these technologies as an entrée into the provision of financial services. Facebook, for example, wants to provide payment services to billions of people using the Libra currency, which relies on distributed ledger technology—although Facebook was recently forced to curtail its ambitions in light of political pressures.

These new technologies could have a big impact in the near future, yet many of them are being developed without any regulatory oversight. Ideally, regulators would oversee the development of new technologies, mandating testing procedures and perhaps requiring that certain features, like circuit breakers, be included. Unfortunately, there has not been any public outcry about fintech’s risks to the financial system, notwithstanding that—as economists Andrei Kirilenko and Andrew Lo have quipped (on the related subject of high-frequency trading—“whatever can go wrong will go wrong faster and bigger when computers are involved.”

A catastrophic fintech failure could cause serious harm to the economy, and we should not wait until it happens to start grappling with the downsides of driverless finance. These technologies are currently in their infancy, but soon they will be established, and unless regulators and policymakers get involved before then, they will have missed their opportunity. If that happens, the public will have been completely unrepresented in the development process—but will still end up footing the bill if the economy blows up as a result of a fintech-generated financial crash.

Future Tense is a partnership of Slate, New America, and Arizona State University that examines emerging technologies, public policy, and society.