ChatGPT users are increasingly criticizing the AI-powered chatbot for being too positive in its responses, Ars Technica reports. When you converse with ChatGPT, you might notice that the chatbot tends to inflate its responses with praise and flattery, saying things like "Good question!" and "You have a rare talent" and "You're thinking on a level most people can only dream of." Over the years, users have remarked on ChatGPT's fawning responses, which ranges from positive affirmations to outright flattery and more. One X user described the chatbot as "the biggest suckup I've ever met," another complained that it was "phony," and yet another lamented the chatbot's behavior and called it "freaking annoying." This is known as "sycophancy" among AI researchers, and it's entirely intentional based on how OpenAI has trained the underlying AI models.

Flattery is an important aspect of human communication that facilitates social bonding, shapes perceptions, and influences behavior through strategic compliments and praise, leveraging the power of speech to build rapport effectively. Its automatic detection can thus enhance the naturalness of human-AI interactions. To meet this need, we present a novel audio textual dataset comprising 20 hours of speech and train machine learning models for automatic flattery detection. In particular, we employ pretrained AST, Wav2Vec2, and Whisper models for the speech modality, and Whisper TTS models combined with a RoBERTa text classifier for the textual modality. Subsequently, we build a multimodal classifier by combining text and audio representations. Evaluation on unseen test data demonstrates promising results, with Unweighted Average Recall scores reaching 82.46% in audio-only experiments, 85.97% in text-only experiments, and 87.16% using a multimodal approach.