With conversational models becoming increasingly available to the general public, developing scalable and robust evaluation metrics is crucial to minimize potential social and psychological risks for the users. Existing evaluation metrics aim to automate offline user evaluation and approximate human judgment of pre-curated dialogs. However, they are limited in their ability to capture subjective perceptions of users who actually interact with the chatbots and might not generalize to real-world settings. To address this limitation, we propose an approach to approximate online human evaluation, leveraging large language models (LLMs) from the GPT-family. We introduce a new Dialog system Evaluation framework based on Prompting (DEP), which enables a fully automatic evaluation pipeline that replicates live user studies and achieves an impressive correlation with human judgment (up to Pearson r=0.95 on a system level). The DEP approach involves collecting synthetic chat logs of evaluated bots with an LLM in the other-play setting, where the LLM is carefully conditioned to follow a specific scenario. We further explore different prompting approaches to produce evaluation scores with the same LLM. The best-performing prompts, which contain few-shot demonstrations and instructions, show outstanding performance on the tested dataset and demonstrate the ability to generalize to other dialog corpora.

Effective question-asking is a crucial component of a successful conversational chatbot. It could help the bots manifest empathy and render the interaction more engaging by demonstrating attention to the speaker’s emotions. However, current dialog generation approaches do not model this subtle emotion regulation technique due to the lack of a taxonomy of questions and their purpose in social chitchat. To address this gap, we have developed an empathetic question taxonomy (EQT), with special attention paid to questions’ ability to capture communicative acts and their emotion-regulation intents. We further design a crowd-sourcing task to annotate a large subset of the EmpatheticDialogues dataset with the established labels. We use the crowd-annotated data to develop automatic labeling tools and produce labels for the whole dataset. Finally, we employ information visualization techniques to summarize co-occurrences of question acts and intents and their role in regulating interlocutor’s emotion. These results reveal important question-asking strategies in social dialogs. The EQT classification scheme can facilitate computational analysis of questions in datasets. More importantly, it can inform future efforts in empathetic question generation using neural or hybrid methods.

Building an empathetic chatbot is an important objective in dialog generation research, with evaluation being one of the most challenging parts. By empathy, we mean the ability to understand and relate to the speakers’ emotions, and respond to them appropriately. Human evaluation has been considered as the current standard for measuring the performance of open-domain empathetic chatbots. However, existing evaluation procedures suffer from a number of limitations we try to address in our current work. In this paper, we describe iEval, a novel interactive evaluation framework where the person chatting with the bots also rates them on different conversational aspects, as well as ranking them, resulting in greater consistency of the scores. We use iEval to benchmark several state-of-the-art empathetic chatbots, allowing us to discover some intricate details in their performance in different emotional contexts. Based on these results, we present key implications for further improvement of such chatbots. To facilitate other researchers using the iEval framework, we will release our dataset consisting of collected chat logs and human scores.