Comparative Analysis of Chatbot Architectures with and without Memory

Memory-enabled chatbots can recall prior conversations, promising more personalized interactions. However, there is little empirical evidence showing whether they consistently outperform non-memory chatbots. This research evaluates whether memory-equipped chatbots provide measurable benefits. We compare a memory-enabled and a non-memory chatbot using ChatGPT on a dataset of 2,000 prompts. Using a 3D evaluation method (Deciphering, Deciding, Describing), prompts are judged across these criteria. Results show both models perform similarly in Deciphering and Deciding, while the memory model performs better in Describing and Essay tasks where personalization improves results.