April 2023 - May 2024
April 2023 - May 2024
Research shows that people over 65 are more resistant to digital banking, primarily due to a distrust of new technology and concerns about artificial intelligence (AI).
This skepticism can lead to two major issues:
1. Older adults might miss out on the benefits of the digital services, resulting in social and economic exclusion.
2. They might over rely on these services, risking financial and emotional harm.
Our project, Conversational Agents for Older Adults (CA4OA), tackles these challenges.
We employ the User-Centred Design (UCD) approach to develop chatbots that are easy to use and build trust among older users.
Our banking chatbot – Fintech Bot – integrates chatbot functionality and simple user interface (UI) elements to assist older adults with their basic banking needs, such as making bank transfers, and includes an AI-powered (large language models, LLMs) chat assistant to engage users in casual conversation with the Fintech Bot, simulating physical bank experience.
Recognising the profound impact of music on mental health, our project has expanded to include a music chatbot specifically designed for older users. Research has consistently demonstrated that music can evoke powerful emotional responses and unlock memories, offering significant benefits such as improved mood and recall of past experiences.
Our Music Bot is designed to be inclusive and emotionally supportive. It engages users by playing songs of their choice and conversing with users their emotional responses to the songs. Playing the role of a music companion, the Music Bot can enhance emotional wellbeing and mitigate the sense of loneliness for some older adults.
By involving elderly users directly in the development process, we can obtain their feedback that shapes our design of the chatbots. This adheres to the principles of UCD, ensuring that the chatbots are functional and meet users’ needs and preference (Figure 1 below). Our overarching goal is to create AI-powered tools that are trustworthy, usable, useful, and desirable for older adults, enhancing their quality of life.
One of the important methods of our research is user-based testing. We have developed two interactive chatbot prototypes: 1. the Fintech Bot (Figure 2) below.
2. The Music Bot (Figure 3) below.
In mid-April, with the University’s ethics approval, we conducted user tests with 16 participants aged from 60 to 81 years old (average= 71).
Eleven participants were female and five were male. The tests took place in-person at Age UK Teesside Head Office, thanks to the generous support of the staff there.
During a testing session, a participant was shown a video demonstrating the banking activities and the Chat feature supported by the Fintech Bot. After watching the video, participants could try out the Bot on a smartphone (Figure 4) for making bank transfers (using fictitious bank account data), checking balances, and viewing account history. For security and privacy concerns, no real personal data were involved.
Using the ‘Chat’ feature, participants can converse with the Bot on any topics except those that are deemed sensitive or privacy/security risky (the LLMs have been configured to filter out inappropriate topics).
After exploring the Fintech Bot, participants were offered the opportunity to test the Music Bot. Five of them gave it a go. Specifically, participants were asked to interact with the Music Bot using music-related queries such as playing a requested piece of music, talking about music genres, artists, and getting music recommendations. After listening to a piece of music, the Music Bot engaged the participant in conversation regarding the music, including their emotional responses (see Figure 3, above).
The majority of participants had an overall positive perception of the user interface of the Fintech Bot. Improvement suggestions were made, for instance slightly increasing the font size. However, trust remains a challenge. A significant observation from the tests was that many users felt trust in the chatbot stemmed largely from branding. In other words, the perceived trustworthiness of the features of the app is essentially determined by the brand’s reputation (i.e. the halo effect). While some participants enjoyed the conversational style and straightforwardness of the Bot, appreciating that it presented only relevant options instead of navigating through extensive and confusing menus, some were sceptical about the necessity of such a chat feature. We aim to address these comments to further improve the chatbot’s design.
Overall, all five participants liked the user interface of the Music Bot and found it interesting to use. They also liked that it could suggest different topics related to music. Participants indicated that they would trust it for listening to music if it played the requested music. We aim to further improve the Music Bot, especially on providing the personalised music recommendations to enhance user trust in it.
Our research journey is ongoing, and our next milestones in the coming months will include:
As trust is highly context-dependent, our research also explores the impact of the application domain on older adults’ trust in chatbots.
While banking is a trust-critical context, we are also investigating the use of chatbot for entertainment purposes (i.e., via a music chatbot), which we found is less trust-critical. In summary, our research project CA4OA is dedicated to enhancing trust and usability in chatbots for older adults, ensuring they can confidently embrace the digital age while staying safe and informed.
This blog was written by Martha T. Correa-Delval, Dr Farkhandah Komal, Effie L-C. Law at Durham University.
You can read the research team’s previous blog Banking Chatbots for You? Hmm… I am still considering… and find out more about their project.