Unremarkable to Remarkable AI Agent: Exploring Boundaries of Agent Intervention for Adults With and Without Cognitive Impairment

Link to paper

Conference: ACM SIGCHI Conference on Computer-Supported Cooperative Work & Social Computing (CSCW), 2025

Authors: Mai Lee Chang, Samantha Reig, Alicia (Hyun Jin) Lee, Anna Huang, Hugo Simão, Nara Han, Neeta M Khanuja, Abdullah Ubed Mohammad Ali, Rebekah Martinez, John Zimmerman, Jodi Forlizzi, Aaron Steinfeld

My Role: Lead Researcher

Research Overview

As the population of older adults increases, there is a growing need for support for them to age in place. This is exacerbated by the growing number of individuals struggling with cognitive decline and shrinking number of youth who provide care for them. Artificially intelligent agents could provide cognitive support to older adults experiencing memory problems, and they could help informal caregivers with coordination tasks. To better understand this possible future, we conducted a speed dating with storyboards study to reveal invisible social boundaries that might keep older adults and their caregivers from accepting and using agents. 

Innovation

Our paper reveals new barriers and opportunities around the adoption and use of agents to help older adults age in place by collaborating with different stakeholders in the care network. Our work searches for boundaries that demarcate agent acceptance at different points along an older adult's aging journey. We compare the perspectives of healthy older adults to those of declining older adults and their primary informal caregivers. Boundaries of agent intervention are dynamic and influenced by changes in the older adult's abilities, especially cognitive. Second, our paper contributes novel implications for agent design and future research through reflections on the findings. Our findings suggest that agents' roles and capabilities need to evolve based on the dynamics of the boundaries. 

Method

The goal of our study is to explore and reveal what might be the boundaries of acceptable and unacceptable agent intervention and how these boundaries may change, starting from when older adults are healthy to when they experience cognitive decline. This includes exploring healthy older adults' anxiety that in the near future they might suffer a decline, along with their anticipation of the adjustments needed after the onset of cognitive decline. We followed a research-through-design approach. 

We chose to conduct a speed dating study using needs validation, which probes desires and values by using storyboards of possible futures. Speed dating as a design research method offers participants many sips of provocative futures using storyboards that illustrate familiar situations with unexpected technology interventions. Our final set of 21 storyboards covered four dimensions : 1) AI agent's autonomy, 2) interpersonal relationships and living situations, 3) older adult's independence and health, and 4) instrumental activities of daily living. 

Participants

We conducted semi-structured interviews with 34 participants across two groups:

• 16 healthy, independent older adults (ages 64-84)

• 9 older adults experiencing early stages of cognitive decline (age 60-83), along with their primary caregivers (age 20-80)

Participants came from diverse backgrounds and low socioeconomic status areas. 

Procedure

Study sessions took place in a private room at the senior center or over video chat via Zoom. Study sessions were audio recorded. Each study session lasted approximately 60 minutes. For each session with healthy older adults, we covered between 2 and 8 storyboards. In two sessions, the participants had to leave early and only reviewed two storyboards. For each session with declining older adults and their caregivers, we covered between 2 and 6 storyboards. Each participant was paid $20.00. Our study was approved by an Institutional Review Board. We transcribed all sessions and analyzed responses using affinity diagramming.

Key Findings

Our analysis revealed  insights around three primary themes: 1) adoption, 2) fears, and 3) boundary of control and unremarkableness. 

1. Paradox of Adoption: Healthy older adults expressed resistance to adopting agent support, while older adults with memory problems and their caregivers thought about using it now and using it in ways that require a long-term understanding of routines and preferences.

2. Fear: Several healthy older adults expressed worries about being a burden to their family and friends and how others might view them. Declining older adults and their caregivers raised concerns about information sharing and privacy. In addition, caregivers conveyed concerns about their caregiving performance.

3. Boundary of Control: Healthy older adults expressed the desire to control the AI agent to operate within a certain space in their care network. Older adults with memory problems want the agent to assert control on their behalf and advocate for them as their own ability to advocate for themselves begins to diminish. 

4. Unremarkableness: Healthy older adults expressed a preference for an unremarkable agent, that is, an agent that exists in the background and is passive.  Older adults suffering from cognitive decline prefer the agent to be remarkable, existing in the foreground and being visible.

Research Impact

This research makes two significant contributions to the field:

• To the best of our knowledge, this concept of agents transitioning from unremarkable to remarkable has not been explored in prior research. Our findings highlight the need for the agent's role to be dynamic throughout different points of the older adult's experience and care network. CSCW and HCI research generally assume that an agent's role is static.

• Healthy older adults expressed resistance to adopting agent support, while older adults with memory problems and their caregivers thought about using it now and using it in ways that require a long-term understanding of routines and preferences. This is a paradox: for the agent to be able to personalize to the declining older adults based on long-term history and understanding, it will need to be adopted before the decline occurs. Our research highlights the need to rethink adoption.

The work has important implications for addressing the growing gap between the population of older adults experiencing cognitive decline and the availability of care workers. By designing agents that can navigate shifting roles appropriately, we may better support aging in place while maintaining trust among all stakeholders.

Future Directions

This research opens several promising avenues for future investigation:

• Adoption: There is an opportunity to rethink adoption in terms of older adults' health stages.  Adoption may need to occur much earlier. It may be beneficial to introduce agents at other times of transition, such as when children start school and their parents begin to encounter complex family logistics around maintaining everyone's increasingly busy schedules---a task for which parents desire technology that can help. Later, if someone in the household starts to experience cognitive decline, the agent can transition to handling more complex social tasks.

• Agent's Role Transitions: What would an agent need to know to become an effective advocate in the care network, and would people listen? 

• Dynamics of Unremarkableness: Where does the appropriate level of unremarkableness lie as the older adult's health declines? How does the change occur? What information can help agents learn to make the transition(s)?

Skills Demonstrated

Technical Skills

• Qualitative research design and implementation

• User experience research methods (speed dating with storyboards)

• Affinity diagramming for thematic analysis

• Audio transcription and qualitative data analysis

Research Skills

• Literature review and synthesis across multidisciplinary fields

• Human subjects research with vulnerable populations

• Semi-structured interview techniques

• Collaborative research in multidisciplinary teams including partnerships with local organizations

• Scientific Communication: Presented complex technical concepts clearly in academic writing and conference presentation

Domain Knowledge

• Aging in place technologies and adoption patterns

• Cognitive decline progression and impacts

• Care coordination and caregiver burden dynamics

• Human-agent interaction design principles

• Ethics in technology for vulnerable populations