Core Concepts
High-ability experts may strategically forego algorithmic decision aids in order to signal their superior diagnostic abilities to consumers and escape a pooling equilibrium with low-ability experts.
Abstract
The content analyzes how technological shocks, specifically the introduction of algorithmic decision aids, affect expert behavior and market efficiency in credence goods markets where experts have heterogeneous and obfuscated diagnostic abilities.
The key findings are:
In the initial phase without investments, experts do not generally maximize consumer income by choosing the optimal price vector. Instead, they often choose the price vector that splits the gains of trade most equally if they behave honestly, leading to efficiency losses.
In the second phase, when experts can invest to improve their diagnostic precision, low-ability experts always invest in the Skill treatment to increase their abilities. High-ability experts, however, are significantly less likely than low-ability experts to invest in the Algorithm treatment, as they can strategically forego the algorithmic decision aid to signal their superior abilities and escape the pooling equilibrium.
Overall, the market is relatively efficient, with high consumer participation rates. However, experts generally under-invest in diagnostic improvements, potentially because consumers do not sufficiently reward beneficial investment patterns.
Investments tend to shift expert price menus towards more self-serving options that reward fraudulent undertreatment, suggesting a potential downside of technological progress in credence goods markets.
Stats
"Experts always choose Pm irrespective of type."
"There is no difference between high-ability expert and low-ability expert behavior."
"Consumers always enter the market."
"Low-ability experts in Skill always invest to improve their diagnostic ability."
"Less high-ability experts than low-ability experts in Algorithm invest to rent the algorithmic decision aid."
Quotes
"High-ability experts may be incentivized to forego the decision aid in order to escape a pooling equilibrium by differentiating themselves from low-ability experts."
"Relying on an algorithmic decision system precludes the expert from sending a competence signal, and may well indicate the opposite."