Designing financial robo-advisers that appeal to people who need them

by Caterina Giannetti, Davide Bacciu, Paolo Crosetto, Alexia Gaudeul, Lorenzo Cominelli, Paolo Pin
Posted on May 15, 2019

The use of artificial intelligence (AI) in robots is an exciting, promising but also slightly frightening area of research. Already, increasingly clever applications help us manage time, make decisions and keep commitments to ourselves and others.

New technologies are also revolutionizing the way we seek and receive advice. For example, applications using AI help consumers correct their eating habits (think of the Weight Watchers app), find a suitable partner (through Tinder, for instance) and adjust their sports training strategies (like the Nike+ running app).

AI in the financial world

AI is shaking up the field of financial advice, where robo-advisers increasingly provide automated and personalized portfolio management.1 Robo-advisers can help people correct for the impact of irrational factors in their decisions. They are particularly attractive because of their low costs, permanent availability and user-friendly interfaces. Robo-advisers also hold the potential to reduce moral hazard problems in the relation between adviser and investor, as they can be designed to unambiguously serve the interests of the investor rather than those of the adviser.2

Existing research suggests that individuals are much more likely to be willing to invest in risky assets if they can obtain financial advice. This is particularly important for investors with little financial knowledge, and for those who are less wealthy, without the luxury of being able to focus on the long term. Only rarely do these investors take advantage of the possibility of earning money by investing some of their wealth.3

Battling the disposition effect

In our project, we set up a series of field experiments to research how robo-advisers can be used to help investors commit to rational investment strategies. We invite our participants to trade on artificial stock markets, accessible through their smartphones, over a period of one month. In the first 15 days, participants trade without a robo-adviser; whereas afterwards, they receive robo-advice on how to trade. Participants are rewarded depending on how well they perform.

The issue we focus on is the disposition effect, which reduces how well people trade financial assets. It is one of the most robust biases found in research literature: stocks that go up are sold too early, and those that go down are sold too late.. Investors subject to the disposition effect may paradoxically attain better outcomes by reducing their own freedom of action. In other words: by committing to trade less often.

Committing our participants

In order to reduce this disposition effect, we give our participants access to a commitment device: an application that gives them trading advice and information. It offers them a way to commit themselves to a particular course of action, helping them to stick to the decisions they made. They may, for instance, commit not to trade for a while and merely observe the evolution of price.

However, investors may not be convinced right away that they actually need a commitment device like the one we are testing. The main objective of our study is to design robo-advisers that are not only good at improving people’s decisions, but that are also likely to be adopted by those who need them most. This is why various participants are exposed to various levels of commitment: some are not allowed to trade at all, others need to pay each time they trade. Some get reminders, others don’t. Which group will do best? We are hoping to find out soon.

1Kaya, O., Schildbach, J., AG, D. B., Schneider, S. (2017). Robo-advice–a true innovation in asset management. Deutsche Bank Research.

2D’Acunto, F., Prabhala, N., Rossi, A. (2018). The promises and pitfalls of robo-advising. Review of Financial Studies.

3Carlin, B., Olafsson, A., & Pagel, M. (2018). FinTech and Consumer Well-Being in the Information Age. Disclosure.

Caterina Giannetti is Assistant Professor at the Department of Economics, University of Pisa, Italy.

Davide Bacciu is Assistant Professor at the Computer Science Department, University of Pisa, Italy.

Lorenzo Cominelli is Biomedical Engineer and Post-Doctoral Researcher in AI for Social Humanoids, Centro Piaggio, Department of Engineering, University of Pisa.

Paolo Crosetto is Researcher (Chargé de Recherche, CR) at INRA, the French National Institute for Agricultural Research, France.

Alexia Gaudeul is Research Associate and Chair of Behavioral Development Economics, Georg-August-Universität Göttingen, Germany.

Paolo Pin is Associate Professor at the Department of Decision Sciences, Bocconi University, Italy.