Ride-hailing drivers, influenced by their contextual experience, exhibit an algorithm aversion preference towards the empty-ride relocating strategy recommended by the platform, which profoundly affects the application effect of algorithm recommendations. This paper presents a simplified network model with two regions to analyze the shared mobility market composed of a ride-hailing platform, passengers requiring cross-region travels, and drivers who independently choose their working areas and relocating strategies. We explore the impacts of the drivers’ algorithm aversion preference on driver behavior, passenger service quality and platform profit. The findings show that when the algorithm aversion preference parameter is low, directly recommending the system-optimal relocating strategy to drivers is the best for platform. However, when the aversion is high, the platform’s optimal recommendation strategy gradually aligns with the drivers’ relocating strategy based on contextual experience. This implies that the platform must proactively “compromise” to make its recommendation more in line with drivers’ contextual experience and cognition to effectively guide their behavior. We also investigate how the driver supply regulates the welfare distribution of algorithm recommendation. It is found that in a moderately supplied market, algorithm recommendation can simultaneously increase platform profit, driver welfare and passenger service quality, achieving a win-win situation for all three parties. In an oversupplied market, although the algorithm recommendation can increase platform profit, it could harm driver welfare by intensifying the competition among them. These findings provide practical insights for the operation of ride-hailing platforms and offer theoretical references for designing a human-machine collaborative decision-making system.