Researchers at Uber are proposing a new artificial intelligence (AI) language model that emphasizes positive, social interaction.
AI is one of the most important developments in the tech industry. Increasingly, is is being used in a wide array of fields, often with the goal to assist individuals in mundane tasks. Chat bots, support agents and conversational AIs are just a few examples. One challenge, however, is making AIs that people will engage with.
Researchers at Uber believe they have the solution, and have written a paper emphasizing the importance of developing an AI language model around positive, social interaction.
Goal-oriented conversational agents are becoming prevalent in our daily lives. For these systems to engage users and achieve their goals, they need to exhibit appropriate social behavior as well as provide informative replies that guide users through tasks.
The researchers hypothesized that an AI using positive interaction would encourage more engagement.
We, therefore, hypothesize that users would prefer a conversaional agent with more polite or more positive language and be more willing to engage with, respond to and persist in the interaction when conversing with an agent using polite and positive language.
Uber’s researchers tested their hypothesis in a ride-sharing environment, where new drivers’ on-boarding was guided by text messages from customer support representatives (CSR).
In this Study 1 we investigated whether and how social language is related to user engagement in task-oriented conversations. We used existing machine learning models to measure politeness and positivity in our analyses. The results show that the politeness level in CSR messages was positively correlated with driver’s responsiveness and completion of their first trip. We also found that positivity positively predicts driver’s first trip, but it has a negative relationship to driver responsiveness even after removing congratulatory milestone messages or messages that do not have any question mark, which usually have positive sentiment and/or do not require responses from drivers.
Uber’s research could be an important stepping stone in the ongoing development of AI, ensuring it best supports human needs.