Consumers feel less social pressure when lying to an AI chatbot, research finds, while human-like social cues reduce dishonest behavior.