Perhaps we need to let go of our expectations of independently functioning machines and instead focus on how we can use machine learning as a human-led tool, Jeffrey Bohn writes in his essay for finews.first.
This article is published on finews.first, a forum for authors specialized in economic and financial topics.
From robo advisors handling customer queries to the gathering of price quotes on comparison websites, Machine Intelligence (MI) has been busily carving out a series of niches for itself in the insurance sector. However, the end result hasn’t always lived up to expectations. Algorithms have more than proven their worth in a range of applications like image sorting or processing complex calculations, leading some to predict their imminent ubiquity. However, this may have been something of a false dawn.
For example, algorithms on comparison sites routinely struggle with the complex nature of what is the best deal to show customers. Basing results on premium figures alone seems simple enough. But once you factor in levels of cover, exclusions, and so on, the waters are soon muddied. Perhaps we need to let go of our expectations and visions of independently functioning machines and instead focus on how we can use machine learning as a human-led tool to enhance the way employees operate and interact with customers.
«In all but the most rudimentary transactions, it’s likely you’ll still want a human presence available»
There’s a host of macro scenarios that MI – particularly the artificial kind – can’t be expected to handle on its own. For example, predicting the chain of events following economic turbulence, or the consequences of political change, is beyond its reach. In fact, any large trend-changing event will likely confuse a machine.
The biggest gains are to be had when we use machine-learning tools to augment the work of people. For example, many businesses have automated much of their marketing, leaving tools and platforms to dictate which ad or article is placed in front of different target audience segments.
However, in all but the most rudimentary transactions, it’s likely you’ll still want a human presence available for a phone call or even a face-to-face meeting at some time during the customer journey to discuss options, personalize a product, or to listen to any lasting concerns. Even more importantly, we still need human experts on call when systems malfunction or underlying environments suddenly change.
«Most of the time, it’s not so much the car that you want to own, it’s the ability to get from A to B that counts»
Self-driving cars are another area where we see success with artificial intelligence. Success in autonomous mobility, however, does not necessarily mean we have adjacent business issues – including insurance – worked out. It’s often said that no one wants to buy an electric drill – it’s the hole in a wall that they’re really after. You could construct a similar argument about cars. Most of the time, it’s not so much the car that you want to own, it’s the ability to get from A to B that counts.
To illustrate the importance of value versus ownership, we've noticed the growing popularity of mobility-as-a-service, facilitated by innovative technologies including all kinds of MI. Consider the following scenario where mobility-as-a-service meets liability.
A customer needs a vehicle, so books one via a mobility-as-a-service platform. The contract exists between those two parties alone, so it’s very straightforward. But as vehicle tech becomes more sophisticated, there’s an almost mind-boggling number of links in the entire value chain. There’s the vehicle manufacturer, the company that wrote the software, the sensor manufacturer, the GPS unit may be from a separate company, too.
«Most cities in the U.S. are reluctant to consent to self-driving cars being tested on the roads»
Now let’s transplant our scenario into an intelligent city and factor in the added complication of the sensors dotted about the city. If there’s a fault in a line of code, or a sensor stops working, or there’s a city-wide system breakdown that compromises the ability of the high-tech vehicle, where does the liability buck stop? We trade the convenience of mobility-as-a-service for increased algorithmic & cyber risks as we become more dependent on sensors, networks, systems, and MI.
Indicative of a fairly conservative approach to the increasing use of technology in cars, most cities in the U.S., for example, are reluctant to consent to self-driving cars being tested on the roads. Yet a lot of that tech has made its way into cars that are already on sale, under the description of driver-assistance aids. For example, the adaptive cruise control that keeps the driver at a safe distance from the car in front. All the new tools available have the potential to make a regular car journey safer and more pleasant, but at least initially, they may also cause momentary confusion and distraction for the driver.
«We’re in something of a grey area or a middle ground»
Increased risks arise because of the hesitancy around the use of driver-tech; we’re not fully embracing it and we’re not completely banning it. Thus, we’re in something of a grey area or a middle ground.
We still have much to do as insurers, software developers, regulators, and cities to figure out how best to design, develop, deploy, & coordinate systems. Machine Intelligence doesn't eliminate risk, rather it transforms the risks we face into ones that may have more appealing characteristics. As a result, human cognition still has a primary part to play in this evolution.
Jeffrey Bohn joined Swiss Re Institute as its Director in 2017, based in Switzerland. He manages all R&D activities across Swiss Re. Prior to his Swiss Re appointment, he served as Chief Science Officer and Head of GX Labs at State Street Global Exchange in San Francisco. Before moving back to California, he established the Portfolio Analytics and Valuation Department within State Street Global Markets Japan in Tokyo. He is fluent in Japanese. He previously ran the Risk and Regulatory Financial Services consulting practice at PWC Japan.
Previous contributions: Rudi Bogni, Peter Kurer, Oliver Berger, Rolf Banz, Dieter Ruloff, Werner Vogt, Walter Wittmann, Alfred Mettler, Peter Hody, Robert Holzach, Craig Murray, David Zollinger, Arthur Bolliger, Beat Kappeler, Chris Rowe, Stefan Gerlach, Marc Lussy, Nuno Fernandes, Richard Egger, Maurice Pedergnana, Marco Bargel, Steve Hanke, Urs Schoettli, Ursula Finsterwald, Stefan Kreuzkamp, Oliver Bussmann, Michael Benz, Peter Hody, Albert Steck, Martin Dahinden, Thomas Fedier, Alfred Mettler, Brigitte Strebel, Peter Hody, Mirjam Staub-Bisang, Nicolas Roth, Thorsten Polleit, Kim Iskyan, Stephen Dover, Denise Kenyon-Rouvinez, Christian Dreyer, Kinan Khadam-Al-Jame, Robert Hemmi, Anton Affentranger, Yves Mirabaud, Katharina Bart, Frédéric Papp, Hans-Martin Kraus, Gerard Guerdat, Mario Bassi, Stephen Thariyan, Dan Steinbock, Rino Borini, Bert Flossbach, Michael Hasenstab, Guido Schilling, Werner E. Rutsch, Dorte Bech Vizard, Adriano B. Lucatelli, Katharina Bart, Maya Bhandari, Jean Tirole, Hans Jakob Roth, Marco Martinelli, Thomas Sutter, Tom King, Werner Peyer, Thomas Kupfer, Peter Kurer, Arturo Bris, Frederic Papp, James Syme, Dennis Larsen, Bernd Kramer, Ralph Ebert, Marionna Wegenstein, Armin Jans, Nicolas Roth, Hans Ulrich Jost, Patrick Hunger, Fabrizio Quirighetti, Claire Shaw, Peter Fanconi, Alex Wolf, Dan Steinbock, Patrick Scheurle, Sandro Occhilupo, Will Ballard, Michael Bornhaeusser, Nicholas Yeo, Claude-Alain Margelisch, Jean-François Hirschel, Jens Pongratz, Samuel Gerber, Philipp Weckherlin, Anne Richards, Antoni Trenchev, Benoit Barbereau, Pascal R. Bersier, Shaul Lifshitz, Klaus Breiner, Ana Botín, Martin Gilbert, Jesper Koll, Ingo Rauser, Carlo Capaul, Claude Baumann, Markus Winkler, Konrad Hummler, Thomas Steinemann, Christina Boeck, Guillaume Compeyron, Miro Zivkovic, Alexander F. Wagner, Eric Heymann, Christoph Sax, Felix Brem, Jochen Moebert, Jacques-Aurélien Marcireau, Peter Hody, Ursula Finsterwald, Claudia Kraaz, Michel Longhini, Stefan Blum, Zsolt Kohalmi, Karin M. Klossek, Nicolas Ramelet, Søren Bjønness, Lamara von Albertini, Andreas Britt, Gilles Prince, Fabrizio Pagani, Darren Willams, Salman Ahmed, Stephane Monier, and Peter van der Welle, Swetha Ramachandran, Teodoro Cocca, Beat Wittmann, Ken Orchard, Michael Welti, Christian Gast, and Didier Saint-Georges.
