How humanizing AI can lead to worse returns for investors
I came across a fascinating study in which humanizing an AI investment advice interface was associated with lower returns for investors.
There are lots of angles to unpack here, but here is a quick, conceptual breakdown of the study and my quick thoughts:
-Participants were given access to a trading platform to monitor their performance
-The availability of AI investment advice was shown to reduce the disposition effect, which was associated with better performance
-However, when the delivery of the AI investment advice was "humanized" (using natural language to chat with the investor, giving the bot a human face and name), investors faired WORSE
-This was a function of the investor asking for advice LESS from the humanized AI advice
This is a striking finding. The very features designed to make these platforms more approachable could push investors away from advice.
This has significant implications when we think about maximizing the outcomes for investors. Investors who select into working with a human advisor are qualitatively different than investors who prefer to manage their own investments. Conversely, someone might choose a DIY platform for a variety of reasons, including not wanting to work with a human or bad past experiences with an advisor, to name just a few.
They don't want a human advisor so humanizing AI tools is not something they want, and moreover, something they might avoid - no matter how good the advice is.
My column on this in The Globe and Mail is unfortunately paywalled, but if you have a subscription click on the article below. The original study is not paywalled, however (you may need a free account with ScienceDirect though): link.
Preet Banerjee
Discussion