Keep Me Updated! Memory Management in Long-term Conversations

Bae, Sanghwan; Kwak, Donghyun; Kang, Soyoung; Lee, Min Young; Kim, Sungdong; Jeong, Yuin; Kim, Hyeri; Lee, Sang‐Woo; Woomyoung, Park,; Sung, Nako

doi:10.18653/v1/2022.findings-emnlp.276

Cited by 3 publications

(12 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Augmenting LLMs to 'remember' past information-often referred to as 'long-term memory' [2,75,77,78,83]-presents significant challenges for two main reasons. First, LLMs can receive input text only within a limited context window (input size).…”

Section: Chatbots Driven By Large Language Models and Long-term Memor...mentioning

confidence: 99%

“…Including the entire conversation session history in the input prompt is thus not feasible for longer-term interactions. One common approach is to include summarized information of the conversation history instead of a raw knowledge base (e.g., [2,41,75]). Second, designing how chatbots should refer to stored information back in conversation involves complex considerations.…”

Section: Chatbots Driven By Large Language Models and Long-term Memor...mentioning

confidence: 99%

“…In this section, we describe the study context of CareCall with LTM as an example of an LLM-driven chatbot deployed to support a public health intervention. This section builds on prior work on the underlying technology (c.f., [2,3,34]) and the design documents written by the designers and developers of CareCall. Building upon previous studies that contributed to the novel implementation of CareCall and LTM, we treat CareCall as a case study for understanding how LTM impacts users' self-disclosure with LLM-based chatbots in the context of public health intervention.…”

Section: Study Context: Clova Carecall and Long-term Memorymentioning

confidence: 99%

“…In this method, a machine learning model generates synthetic dialogues from a small set of human-written dialogues and trained workers flag and fix errors in the synthetic dataset (please see [3] for more details about the data augmentation technique). The example dialogues, which consist of dialogues relevant to five health topics (meals, sleep, health, going out, and physical activity), steer the LLM to generate appropriate responses in the public health intervention context while aiming to avoid selecting undesirable or potentially harmful phrases [2].…”

Section: Carecall and Long-term Memorymentioning

confidence: 99%

“…LTM was incorporated into CareCall to improve its ability to provide emotional support by offering familiarity with users. At the end of each session, a summarizer driven by an LLM ( F ○ in Figure 1b; [2]) generates summary sentences that are relevant to the five LTM topics (Section 3.2). The memory management layer ( D ○ in Figure 1b) stores and updates the summary sentences upon each call (e.g., Removing the "Regular visit to a clinic due to leg pain" status after a user reports that they have completed the treatment).…”

Section: Carecall and Long-term Memorymentioning

confidence: 99%

See 4 more Smart Citations

Understanding the Impact of Long-Term Memory on Self-Disclosure with Large Language Model-Driven Chatbots for Public Health Intervention

Jo,

Jeong,

Park

et al. 2024

Proceedings of the CHI Conference on Human Factors in Computing Systems

View full text Add to dashboard Cite

Recent large language models (LLMs) offer the potential to support public health monitoring by facilitating health disclosure through open-ended conversations but rarely preserve the knowledge gained about individuals across repeated interactions. Augmenting LLMs with long-term memory (LTM) presents an opportunity to improve engagement and self-disclosure, but we lack an understanding of how LTM impacts people's interaction with LLM-driven chatbots in public health interventions. We examine the case of CareCallan LLM-driven voice chatbot with LTM-through the analysis of 1,252 call logs and interviews with nine users. We found that LTM enhanced health disclosure and fostered positive perceptions of the chatbot by offering familiarity. However, we also observed challenges in promoting self-disclosure through LTM, particularly around addressing chronic health conditions and privacy concerns. We discuss considerations for LTM integration in LLM-driven chatbots for public health monitoring, including carefully deciding what topics need to be remembered in light of public health goals. CCS CONCEPTS• Human-centered computing → Empirical studies in HCI; Natural language interfaces; • Computing methodologies → Natural language generation.

show abstract

Section: Chatbots Driven By Large Language Models and Long-term Memor...mentioning

confidence: 99%

Section: Chatbots Driven By Large Language Models and Long-term Memor...mentioning

confidence: 99%

Section: Study Context: Clova Carecall and Long-term Memorymentioning

confidence: 99%

Section: Carecall and Long-term Memorymentioning

confidence: 99%

Section: Carecall and Long-term Memorymentioning

confidence: 99%

See 3 more Smart Citations

Understanding the Impact of Long-Term Memory on Self-Disclosure with Large Language Model-Driven Chatbots for Public Health Intervention

Jo,

Jeong,

Park

et al. 2024

Proceedings of the CHI Conference on Human Factors in Computing Systems

View full text Add to dashboard Cite

show abstract

Understanding the Benefits and Challenges of Deploying Conversational AI Leveraging Large Language Models for Public Health Intervention

Epstein

Jung

et al. 2023

Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems

View full text Add to dashboard Cite

Persona-Identified Chatbot through Small-Scale Modeling and Data Transformation

Keum,

Sun,

Lee

et al. 2024

Electronics

View full text Add to dashboard Cite

Research on chatbots aimed at facilitating more natural and engaging conversations is actively underway. With the growing recognition of the significance of personas in this context, persona-based conversational research is gaining prominence. Despite the abundance of publicly available chit-chat datasets, persona-based chat datasets remain scarce, primarily due to the higher associated costs. Consequently, we propose a methodology for transforming extensive chit-chat datasets into persona-based chat datasets. Simultaneously, we propose a model adept at effectively incorporating personas into responses, even with a constrained number of parameters. This model can discern the most relevant information from persona memory without resorting to a retrieval model. Furthermore, it makes decisions regarding whether to reference the memory, thereby enhancing the interpretability of the model’s judgments. Our CC2PC framework demonstrates superior performance in both automatic and LLM evaluations when compared to high-cost persona-based chat dataset. Additionally, experimental results on the proposed model indicate the improved persona-based response capabilities.

show abstract

Keep Me Updated! Memory Management in Long-term Conversations

Cited by 3 publications

References 0 publications

Understanding the Impact of Long-Term Memory on Self-Disclosure with Large Language Model-Driven Chatbots for Public Health Intervention

Understanding the Impact of Long-Term Memory on Self-Disclosure with Large Language Model-Driven Chatbots for Public Health Intervention

Understanding the Benefits and Challenges of Deploying Conversational AI Leveraging Large Language Models for Public Health Intervention

Persona-Identified Chatbot through Small-Scale Modeling and Data Transformation

Contact Info

Product

Resources

About