Digital learning, now with robot voices, chatbots, and AI girlfriends

I remember the first time I heard ChatGPT speak. It reminded me of a scene from Halt and Catch Fire: the Apple Lisa boots up and announces, “Hello. I am Macintosh.” The moment is uncanny not because the machine talks, but because it claims presence when it does.

This raises a basic question for education: do our learning environments speak with our voices? And, just as importantly, do they listen?

Beyond topic structures, many digital learning environments remain essentially mute. Their architecture offers content and teaching presence is inferred rather than experienced. Guidance is assumed rather than articulated. Listening is even rarer. Where, in most digital curricula, are learners actually heard? End-of-module surveys? Occasional assessment feedback? Perhaps a discussion forum that rarely closes the loop. These are not dialogue; they are administrative echoes.

Outside education, however, our digital ecosystem listens constantly. Commercial UX and platform design have evolved into a sophisticated language—one many people now acquire from early childhood. I have watched a one-year-old swipe through a phone’s photo gallery, pausing and cheering when it recognises itself (or perhaps another baby). These systems reduce friction, anticipate behaviour, allow error and correction, and continuously nudge attention. They suggest and call to action, and seduce participation. They dialogue fluently.

Over the past eighteen months, I have been asked repeatedly to advise on automated AI marking, AI-narrated videos generated from PowerPoint slides and the integration of different AI features and tools. These requests usually combine two pressures: strategic aspirations around quality and consistency, and an environment of acute scarcity—time, skills, staffing. There is no argument here of adopting AI confidently, as skilled, critical professionals; we should. The matter is that the proposed request is often voice without presence.

And voice is not neutral. Voice establishes a relational bond. It invokes authority, after all, in narrative, gnarus is the one who knows and is listened to. Voice contextualises, guides, reassures, questions. Online, voice replaces physical co-presence to a great extent. It threads activities together, explains purpose, motivates effort, frames assessment, and establishes safety. Voice mediates meaning. It selects and emphasises. It orders experience. It motivates action. It is both relational and functional.

In virtual learning environments, voice is present—or absent—everywhere: in instructions, learning outcomes, scaffolding, workload guidance, assessment explanations, feedback, announcements, profile information, facilitation, rubrics, completion rules, and welcome videos. Silence here is not neutral; it transfers labour to students. Dialogue goes further. In digital learning, dialogue can be Socratic, didactic, or dialectical. It takes shape through discussion, collaboration, debate, and assessment loops. 

Dialogue is more than the sum of individual voices—it is relational engagement. In asynchronous learning, two thirds of meaningful engagement occur through relationships: with peers and with educators. Remove dialogue, and what remains is task completion. What is largely missing is guided independent study with rich asynchronous feedback—the kind that sustains engagement without physical co-presence. The kind that allows trust, storytelling, identity, and Socratic questioning to emerge over time. In other words, the conditions for humanity in asynchronous learning.

This invites a harder question.

Over the past two decades, as our digital engagement as citizens exploded, our educational digital spaces remained content-driven and uninhabited. Where did students learn the skills that sit at the intersection of digital communication, wellbeing, and citizenship? Where did they practise critical listening, rhetorical judgement, disagreement, and deliberation?

They learned them elsewhere.

They learned them in platforms where discourse accelerated, identity hardened, persuasion became algorithmic, and reflection was structurally disincentivised. Spaces where turn-taking collapsed, emotional appeal trumped evidence, and immediacy displaced deliberation. Where speed compressed temporal distance and shifted communication from rhetoric to reflex, intentionally.

These platforms privilege moral outrage, threatened with loss aversion, continually relied on in-group signalling, and narrative immersion. They reward conformity over nuance. They produce affective architectures in which responses become automated—not only by machines, but by us.

As educators, and especially as digital educators, what large-scale alternatives did we build? Where, in our curricula and platforms, did we design for slow dialogue, critical listening, lateral reading, refutation of premises, ethical argumentation, and respectful disagreement? I know what happened in the few cases we did, but was this worth fighting for harder? Where did we formally assess these capacities? Where did we make space for shared, co-created axiological depth?

The deeper risk here is ontological.

When we consider this risk of a greater integration of GenAI into our practices, professional identities and digital learning ecosystems, it helps to look outward—into the digital territories where people now spend the most time and emotional energy. These are the spaces that have most powerfully shaped sexuality, intimacy, identity, and politics over the past two decades. They are not peripheral to education; they are where learning about relationship, desire, persuasion, and belonging has already been taking place.

How were these spaces shaped? Did they priviledge voices in dialogue—listening, responding, shaping meaning together? Or did we simply add attention, need, and affective energy into systems optimised to extract and redirect it? Influence is the core goal of communication, gained through attention; it has always been.

Extreme cases of engagement with what were supposed to be production and mediation technologies (porn addiction, AI therapy and romatic/sexual relationships) are not curiosities at the margins; they are extreme but clarifying examples of relational replacement. They show what happens when technologies designed to mediate between people instead become substitutes for relationship itself. Where interaction remains responsive but is no longer reciprocal. Where presence is simulated, but vulnerability, mutual transformation, and ethical friction are absent.

These systems listen. They respond. They adapt. But they do not engage in dialogue. They do not resist, challenge, or co-create meaning; they don't participate in the axiological, embodied, reciprocal, historical relationship we have with the world and with each other, the fruit of which shapes everything from how we seek pleasure into how we write law, agree on ethics, live religion. These systems absorb desire, mirror identity, and reward projection. They replace the risks of relationship with the safety of control. And then isolate and alienate. 

This is the territory into which education is now moving its voice.

The risk, then, is not simply that AI tools are integrated badly, or without sufficient ethical guidance; we should integrate them as confident producers, as creators, as skilled professionals. The risk is that—having left our asynchronous learning spaces largely uninhabited—we allow voice, presence, and narrative authority to be delegated to systems whose primary competence is not education, but affective capture, done by this demand for lower effort and higher productivity. 

If we are not careful, as with previous media technologies, the shift will be subtle but decisive: from how we communicate to who or what is communicating. From mediation to replacement.

If we do not intentionally reclaim voice and dialogue in digital learning—human voice, human listening, human disagreement—others will speak in our place. And they will speak fluently, continuously, and without the friction that education, by necessity, must preserve. The friction - multiple perspectives, disagreement, turn-taking, negotiation - which is how we learn not only to be with the Other, and to grow as ourselves, but to continuously co-create, negotiate, experience and update our values.

We did  it wrongly before, shifting work to students, from poor asynchronous learning and assessment preparation and assessment design, ignoring cultural changes and favouring scale and cost over the demands of a new set of capabilities and values. Now,  we risk signalling, with this relational break, that we value something above dialogue, something which mimics it only; and students will, as they have already, do the same.