This thesis explores the possibility of incorporating social robots equipped with LLMs to assist geriatric nurses in strengthening social interaction and enhancing care for the elderly. The main goal is to investigate how social robots, with advanced LLM capabilities, can be taught to demonstrate social abilities similar to those of human caretakers, such as persuasiveness, which is crucial in everyday eldercare tasks. The project seeks to establish specific principles for these robots, with a particular emphasis on their capacity to effectively involve elderly in social activities.

The study employs a mixed-methods approach, incorporating qualitative interviews with nurses and a quantitative simulation experiment to assess the effectiveness of social robots utilizing an LLM, notably GPT-4. The qualitative component is conducting interviews with healthcare professionals to get insights into the existing difficulties and needs in eldercare, with a particular focus on the significance of empathy, trust-building, and tailored communication.

The interviews demonstrate that trust and customized interactions are crucial elements in convincing elderly folks to participate in activities, underscoring the importance of social robots continually displaying these attributes. During the quantitative phase, a simulation experiment is carried out to evaluate the efficacy of social robots in convincing elderly to engage in particular activities, such as taking a stroll. The experiment entails fabricating fictional synthetic users with diverse backgrounds and assessing the influence of increasing degrees of robot persuasiveness on user adherence. The results of the simulation highlight the significance of maintaining a balance between persuasiveness and empathy in order to attain maximum engagement and cooperation from the elderly. An excessive level of persuasiveness might provoke resistance, whilst an insufficient level of persuasiveness can lead to a lack of motivation. Furthermore, the study identifies various crucial factors that have an impact on convincing elderly. These factors include the need for tailored and situation-sensitive communication, the creation of trust through consistent and dependable interactions, and the significance of emotional intelligence in establishing a connection with users. These criteria are utilized to establish precise benchmarks for training social robots, guaranteeing their ability to adjust their communication style effectively to cater to the unique requirements and preferences of elderly. Furthermore, the thesis examines the wider consequences of incorporating LLMs into social robots for the purpose of providing care for the elderly. These technologies have the ability to reduce the workload of human caregivers by offering individualized, empathetic, and efficient communication that improves the quality of care. Nevertheless, the research also recognizes the constraints of existing LLM technology, including difficulties in comprehending subtle human emotions and sustaining reliable, contextually aware interactions. It highlights the necessity for additional progress in LLMs to tackle these difficulties and enhance their suitability in eldercare environments. Moreover, the thesis explores the ethical and sociological consequences of employing social robots in eldercare, specifically with matters of privacy, autonomy, and the risk of excessive dependence on technology. A balanced approach is necessary to take into account the advantages of social robots while also ensuring that they enhance rather than substitute human caretakers, thus safeguarding the dignity and independence of the elderly. Ultimately, this thesis examines how social robots equipped with integrated LLMs can improve eldercare by assisting geriatric nurses in fostering social interaction and enhancing the level of care for the elderly. The results underscore the significance of providing these robots with instruction in crucial social abilities, such as empathy, trust development, and tailored communication, in order to guarantee their successful incorporation into eldercare settings. The study emphasizes the importance of continuous advancement and ethical deliberation in the implementation of social robots, with the goal of establishing a nurturing and enhancing setting for the elderly. Future research should prioritize conducting longitudinal studies to assess the enduring effects of social robots in eldercare and investigate other improvements in LLM technology to augment their capacities and efficacy.