(1) Yeonheebot: Mental Health Care Conversational Agent for the Elderly
Ryu, H., Kim, S., Kim, D., Han, S., Lee, K., and Kang, Y. 2020. Simple and Steady Win the Healthy Mentality: Designing a Chatbot Service for the Elderly. In Proceedings of the ACM on Human-Computer Interaction, Vol. 4, CSCW2, Article 152 (October 2020), 25 pages, https://doi.org/10.1145/3415223
- Tools Used: Kakao Talk I Developer, Raspberry Pi server
Although suicide among older adults is increasing in severity in the aging among South Koreans, older adult depression and anxiety, which are two crucial factors affecting suicide in this population, are not adequately addressed by the status quo. Existing products and services emphasize physical needs; however, they still lack a user-oriented approach. Moreover, offers for mental health care are costly and exclude potential users from the development process, thus inhibiting older adults from having better access to such services. To address this problem, we designed and developed a mental health care chatbot, Yeonheebot, to decrease anxiety and depression among older adults. Yeonheebot initiates conversations regarding daily care for the well-being of older adults and provides features related to their interests as initiatives for using the mental health care chatbot. To examine whether a short-term, yet repeating conversation with the agent could help reduce anxiety and depression among older adults, we conducted a 2-week field study with 25 users. The results indicate that the constant use of Yeonheebot decreased the levels of anxiety and depression by 36% and 18%, respectively. Our research implies that simple and repeated interactions could help Korean older adults cope with anxiety and depression.
(2) Carrie the Carebot: Chatbot for Dementia Caregivers
Launched on expertisecentrum dementie vlaanderen official homepage
- Key Research Question: What is the extent to which chatbots can assist informal dementia caregivers?
- Tools Used: Microsoft Azure Framework V4 SDK, QnA Maker, Landbot.io
To solve the imbalance between the exponentially growing informational demand from informal caregivers of people with dementia and the slowly increasing supply of informational provision centers, we believe that chatbots are the ideal medium to fill such a gap. Under this premise, the research probed the context and personality that the informational caregivers would be expecting from an informational chatbot. As for the interface of the chatbot, the preference of the friendly persona was proven to be true and users were seeking an easier way of navigation than solely text-based navigation. In the case of the context, although the four inquiry categories defined from the crawling of frequently asked questions on the web are deemed as crucial in informational provision, the interviews indicated regional differences on the emphasis of each of the categories. Thus, our hypothesis that the chatbot should be built upon the database emphasizing the caregiving methods was proven wrong. However, the actual usage of the chatbot during the two-week launch on the official homepage of Expertisecentrum Dementie Vlaanderen showed that the preconception of what people expect from chatbots can differentiate their inquiries from when they ask questions to human experts in that they are more likely to ask general questions rather than questions that need more personal input such as explaining the problem that they are experiencing with the chatbot extensively. Based on these findings, Carrie the Carebot can be further developed to better meet the general needs of informal caregivers of dementia and alleviate the burdens that they face.
(3) UICbot: Chatbot for Reducing Administrative Workers' Burden
1st Place in Chatbot Service Design Competition
- Key Research Question: How can chatbots reduce administrative workers' burden?
- Tools Used: Flask, KakaoTalk RestfulAPI, SQLite
UIC chatbot has been developed to answer the students’ inquiries with accuracy and specificity and reduce the heavy and redundant workload for administrative officers. It could be considered as an integrated question and answer assistant for it has incorporated the dispersed information into a single database with precise categorization. The aforementioned precise categorization is taken into effect by including both category buttons and free-format dialogue into the chatbot. Thus, its users could benefit from saving the time to search for the answers to such time-consuming questions on their own and from obtaining the more accurate response to what they have inquired.
The UIC chatbot framework was built using Flask and uses KakaoTalk Plus Friend as its platform. Since UIC chatbot uses KakaoTalk Plus Friend as its platform, it is accessible to even the most technologically ignorant users in regard to the fact that KakaoTalk is the most prevalent messenger platform in Korea. The chatbot includes important information that students need such as regulations and add and drop periods. A question that can be answered in a succinct manner is provided with an answer while a question that needs a lengthier answer is connected with the relevant url for students’ information.
The UIC chatbot provides answers to the questions in various ways to provide updated and reliable answers: real time web crawling, GitHub and database. ‘Yonsei notice’ and ‘UIC’ websites are crawled in real time to provide students with updated information such as student service, academic affair and career information. The degree requirement is uploaded as a pdf file in GitHub and connected with the chatbot. This allows the students to search their degree requirements according to their class and major in a much faster manner.