Algorithm Reliance Under Pressure : The Effect of Customer Load on Service Workers

The algorithm-augmented business model promises service companies the benefits of both algorithms (quality and efficiency) and humans (human touch). However, companies will only realize this promise if their customer-facing workers rely on the algorithms, and existing research presents conflicting arguments about workers’ willingness to do this. People are often unwilling to rely on algorithms for forecasting and prediction, but might be more willing to do so in the service setting because their customer loads create pressure to work efficiently. We design a laboratory experiment in collaboration with ReUp Education, an algorithm-augmented success coaching company, to resolve this conflict. We situate participants as service workers making personalized recommendations—with algorithmic support—to a series of customers. We first study the effect of customer load on algorithm reliance. We find that although algorithm reliance improves performance, participants generally rely very little on algorithms. However, higher customer loads increase algorithm reliance. The effect of customer load on reliance grows over time as the high customer load creates more opportunities for experience with, and feedback about, the algorithm. We then test the effect of learning interventions on algorithm reliance. We find that participants can be encouraged to rely more on algorithms through learning interventions, especially interventions that present abstract descriptions about the algorithm’s benefits. Our results contribute to theory at the intersection of service operations and algorithm reliance and have practical implications about the algorithm-augmented service model