ONGOING RESEARCH
When AI Gets You Wrong
Personalisation, lifelong learning and older adults’ engagement with generative AI
Examines what happens when generative AI personalisation gets older learners wrong, and how these mismatches affect their ability and willingness to keep learning with AI.
Research lead:
Xuan Thanh Nguyen
Research Lead:
Key takeaway
Personalised AI does not necessarily understand the person it is responding to.
Older adults may need to correct, verify or work around AI responses that do not fit their circumstances or goals.
The research could inform more adaptable, context-sensitive and age-respectful AI-supported services.
Impact in practice
The research could help AI developers, service designers, educators and policymakers understand where AI personalisation fails and design services that better reflect the diversity of older adults' circumstances, goals and learning needs.
About the research
Generative AI can adapt explanations, make recommendations and respond conversationally, but its apparent personalisation does not always reflect a genuine understanding of the person using it.
When AI Gets You Wrong examines what happens when AI-supported learning does not fit a person’s circumstances, needs, knowledge or goals.
Focusing on adults aged 60+ who use generative AI for lifelong learning, the project investigates how these forms of personalisation misalignment arise, how older adults recognise and respond to them, and how they affect continued engagement with AI.
Why it matters
Generative AI could make lifelong learning more accessible by helping people ask questions, explore interests and receive tailored explanations. However, it may also make inappropriate assumptions, provide unsuitable advice or produce information that users must correct and verify.
For older learners, these mismatches can create additional work and make AI-supported learning less effective. Understanding these experiences can help ensure that AI systems respond to the diversity of older adults’ goals, circumstances and support needs rather than treating later-life users as a homogeneous group.
The project will contribute a consumer-oriented definition and classification of generative AI personalisation misalignment, evidence about how older adults experience and respond to these mismatches, and insights for designing more adaptable, context-sensitive and inclusive AI-supported learning services.
Research focus
The central question guiding the project is "How does generative AI personalisation misalignment shape older adults’ engagement in lifelong learning?"
The research explores:
Personalisation misalignment: the different ways apparently personalised AI responses can fail to fit the person receiving them.
Older adults’ experiences: how older learners recognise, interpret and experience these mismatches.
Responding to AI: how people correct, verify, challenge, adapt to or work around inappropriate AI responses.
Learning engagement: how different forms of misalignment relate to people’s willingness and ability to continue learning with AI.
Diversity in later life: how people’s different goals, experiences, circumstances and support needs shape their interactions with AI.
Inclusive AI design: what these experiences can teach us about creating more context-sensitive and age-respectful AI services.
How the research is being conducted
When AI Gets You Wrong uses a multi-stage research design that brings together conceptual development, co-design, observed interactions and quantitative research.
The project begins with a systematic review of existing research to develop a definition and initial classification of different forms of generative AI personalisation misalignment.
The project is now beginning data co-design of the research instruments. Older adults will contribute to the development and refinement of the learning activities, interview materials and other research procedures used in subsequent stages. This process will help ensure that the instruments are understandable, relevant and appropriate for the experiences of older learners.
The qualitative research will examine people’s experiences while they interact with generative AI. Methods include interviews, learning tasks, observation, stimulated recall, selective think-aloud and, where approved, screen recording. This allows the research to examine not only what people say about AI afterwards, but also what happens when a mismatch occurs during an interaction.
A survey of approximately 590–600 adults aged 60+ who have recently used generative AI for learning will then examine how different forms of personalisation misalignment are associated with learning engagement.
Together, these stages connect what personalisation misalignment is, how it is experienced, what people do when it occurs, and how it relates to continued engagement with AI.
Meet the Researcher
Xuan Thanh Nguyen
Xuan Thanh Nguyen is a PhD candidate at RMIT University whose research explores how generative AI personalisation affects older adults’ engagement in lifelong learning. His work examines where personalised AI support can fail to reflect people’s lived contexts and how more inclusive, age-respectful and context-sensitive AI services can be designed.
Who the research is for
The research involves adults aged 60+ who use generative AI for learning, recognising the diversity of older adults’ experiences, capabilities, interests and reasons for learning.
The findings are also relevant to adult-learning providers, community organisations, educators, AI and service designers, technology developers and policymakers seeking to make AI-supported services more inclusive.
Potential impact
When AI Gets You Wrong aims to support a more inclusive understanding of AI personalisation.
The research will develop a consumer-oriented definition and classification of generative AI personalisation misalignment, helping researchers and practitioners identify how apparently personalised AI can fail to fit people’s circumstances.
It will also provide evidence about how older adults recognise, interpret and respond to these mismatches, and whether different forms of misalignment are associated with their engagement in AI-supported learning.
For technology and service designers, the findings can inform AI interactions that are more adaptable, context-sensitive and responsive to diverse users. For educators, community organisations and policymakers, they can support older adults’ engagement with AI without assuming that difficulties arise simply because of age or limited digital capability.
Ultimately, the project asks "if AI is going to personalise our experiences, how can we make sure that personalisation actually fits the people it is intended to support?."