Source-linked AI summary
An emancipatory vision for designing (generative) AI for learner flourishing
Luis P. Prieto, Yannis Dimitriadis
TL;DR
GenAI’s excessive adoption coincides with concerns about learner over-reliance and isolation, while current AIED proposals rest on assumptions the paper considers problematic. The paper presents an emancipatory vision with tentative principles and a design methodology; it remains embryonic, with important methods, exemplars, and measures still unresolved.
Problem
GenAI use raises concerns, while current AIED proposals rely on problematic assumptions about agency, predictable effects, permanent scaffolding, values, and individualistic learning.
Source: §1 · pp. 1–2Method
Results
The paper proposes orienting educational AI toward long-term learning, learner agency, and flourishing within nested complex systems.
Source: §1 · p. 2Takeaways & Limitations
The proposed process evaluates diverse outcomes over time and across social levels, iteratively revising designs and modeling wider-scale diffusion and appropriation.
Source: §4 · p. 4The vision remains embryonic, with concrete design methods, exemplars, and measures of social and environmental constructs still open challenges.
Source: §5 · p. 5
Abstract
from arXiv · showhide
The hype around generative AI seems to promise unprecedented productivity (and learning) gains. However, these technologies' increasing agentic features seem to push learners towards individualism (or individual isolation), over-reliance, and dependence on them. Human-centered design approaches (e.g., value-sensitive design) assume that, by unearthing human needs, preferences, and values, technology researchers/designers may avoid such dangers, which are driven by wider systemic factors like economic incentives or inherent human limitations (e.g., our tendency to seek, in the moment, the easiest path of action). Yet, so far these efforts seem insufficient to guide our design of educational technology that avoids the aforementioned dependency and isolation dangers, while finding widespread adoption. This paper presents an alternative, more emancipatory vision for future educational AI technology, oriented towards learner flourishing while considering the wider complex systems they inhabit, including tentative design principles and an overall design methodology. Yet, many open questions remain before this vision can be realized.
1. Introduction
GenAI has seen what the authors call “excessive adoption,” alongside concerns about over-reliance, metacognitive laziness, social isolation, and well-being. The paper argues that problematic assumptions in current AIED design and gaps in value-sensitive design call for a more emancipatory vision focused on learner flourishing.
1. Introduction
- GenAI’s “excessive adoption” accompanies concerns about over-reliance, metacognitive laziness, social isolation and lower well-being, and environmental impacts.The authors also note that workplace productivity gains remain unclear.Source: §1 · p. 1
- Current AIED proposals rely on problematic assumptions about tool agency, designers’ ability to predict effects, permanent scaffolding, value choices, and individualistic learning.The assumptions also include overlooking the heterogeneous tool ecology and resource constraints.Source: §1 · pp. 1–2
- These assumptions risk learner agency and flourishing, and may leave learners socially isolated and dependent on educational technologies and their owners.The paper also identifies risks to societal and environmental flourishing.Source: §1 · p. 2
- Value-sensitive design addresses some concerns but retains assumptions about predictable outcomes and simplified values, and offers no specific guidance for education.The paper notes that resolving value tensions remains an unsolved challenge.Source: §1 · p. 2
- The paper proposes an emancipatory AIED vision for long-term learning, agency, and learner flourishing, with tentative design principles and a general methodology.It also outlines open questions and dilemmas for realizing the vision.Source: §1 · p. 2
2. A more emancipatory vision: AI for learner flourishing
The proposed vision redirects AIED from increasingly capable, agentic systems toward learner agency and flourishing within complex educational systems. It calls for humility about deployment, limited and temporary scaffolding, attention to tensions among values, and recognition of education as communal socialization.
2. A more emancipatory vision: AI for learner flourishing
- AIED should optimize for learners’ future agency and flourishing rather than ever more capable and agentic tools.Learner agency is described as the capacity to act intelligently and flexibly in the future.Source: §2 · p. 2
- Designers should treat deployment as the beginning of appropriation and “unfolding”, or “design-in-use”, rather than assume they can predict effects.The vision emphasizes intellectual humility about the technology ecosystem and second-order effects.Source: §2 · p. 2
- AIED should intervene only as much and as long as needed, using minimal resources so learners can outgrow and leave its scaffolding.Source: §2 · p. 2
- AIED should navigate tensions within learners, among stakeholders, and across societal and environmental levels, while treating education as communal socialization.Learning involves individual reflection as well as interaction with peers, teachers, society, and the environment.Source: §2 · p. 3
- The authors expect this approach to help learners face challenges and uncertainty, but say societal and environmental flourishing requires critical evaluation.The proposed focus includes awareness, connection, insight, and purpose.Source: §2 · p. 3
3. Tentative design principles
The paper tentatively proposes ten principles for AIED, emphasizing learner agency, adaptable and temporary support, social connection, plural values, flexibility, metacognition, and interoperability.
3. Tentative design principles
- Design tools and services that support multi-level learner agency, are co-created through appropriation, and fade as learners become more autonomous.The paper favors under-designed or adaptable systems and observing how stakeholders repurpose them in practice.Source: §3 · p. 3
- Design to hand learners’ attention from the screen to real-world collaboration and environments, while avoiding systems that pose as people or foster inappropriate attachment.The authors warn that LLMs’ human-like language and emotional expression may enable “attachment hacking.”Source: §3 · p. 3
- Support multiple theories and values rather than one outcome metric, potentially through modular systems that combine approaches for learners and their social and environmental contexts.Source: §3 · pp. 3–4
- Keep designs minimal and flexible under uncertainty, and foster learner resourcefulness, adaptability, resilience, and metacognition.The paper connects resourcefulness with Inuit qanuqtuurniq and metacognition with deep learning and flourishing.Source: §3 · p. 4
- Make systems interoperable with existing educational practices and technologies through options such as open formats, import/export, and configurable goals and values.Source: §3 · p. 4
4. An overall design methodology
The proposed design methodology combines contextual inquiry and stakeholder values with minimal interventions, authentic deployment, multi-level evaluation, and iterative scaling. It also calls for modeling wider diffusion and appropriation, then sunsetting projects when their ripple effects can no longer be measured.
4. An overall design methodology
- Elicit diverse stakeholder values and examine existing educational and social practices through meaningful contextual inquiry.The inquiry need not be overly long or laborious.Source: §4 · p. 4
- Create a minimal viable prototype that supports the target learning phenomenon and learner flourishing within the existing socio-technical system.Minimalism includes system complexity and resource needs, not only the interface.Source: §4 · p. 4
- Deploy in authentic settings and assess multiple outcomes at social levels and timeframes, including how people appropriate or repurpose the technology.Outcomes include cognition, metacognitive ability, and well-being, measured immediately and after delays where possible.Source: §4 · p. 4
- Iteratively revise toward “good enough” results across levels and outcomes, scale deployment, model wider diffusion and appropriation, and sunset projects when ripple effects are no longer measurable.Source: §4 · pp. 4–5
5. Open questions and conclusion
The paper does not reject GenAI: it identifies properties that could support its vision, while emphasizing that the vision remains embryonic and requires further community work.
5. Open questions and conclusion
- GenAI’s language, democratizing, and concept-fusing properties may support reflection, stakeholder participation, and theoretical/value syncretism without requiring ever-larger models.The authors suggest asking what smallest LLM can provide the needed functions and properties.Source: §5 · p. 5
- A locally run, comparatively small open model is presented as a likely way to limit ecological impacts compared with a resource-intensive remote data center.Source: §5 · p. 5
- Realizing the vision requires concrete methods for resolving value tensions, exemplar AIED systems and evaluations, and practical measures of group- and environment-level constructs.The authors also identify design tensions in relying on LLMs developed by foreign companies, including their biases and incentives.Source: §5 · p. 5
- The vision remains embryonic and depends on extensive discussion and joint work among researchers and designers to move beyond optimizing for one learner toward societal and environmental flourishing.Source: §5 · p. 5