Independent · Education

Learning with AI in 2045

Imagining embodied AI beyond chatbots.

Overview

AI saves effort, learning needs effort, and context decides how much to keep.

SXSW EDU 2027 · Workshop · Austin, March 13–16 ↗

A question gets a quick AI answer, which keeps understanding lower; a learner moves through the effort of learning to understanding, past a dial for context, to forms of AI beyond the chatbot: an object, a ritual, a companion, and an open question
Role
Co-lead researcher and workshop designer
Timeline
2026, ongoing
Teams
Co-led with Sneha Arvind

Context

AI is becoming an ordinary part of learning. Most of these tools follow UX heuristics: reduce memory load, cut steps, keep things effortless. Yet cognitive effort, or productive friction, is often where deep understanding forms, and removing it by default may strip away the conditions learning needs. Early research links this to cognitive de-skilling, and overreliance, overconfidence, and weaker independent performance are emerging as costs of frictionless AI.

Sources: Lee et al., Microsoft Research, CHI 2025 · Kosmyna et al., MIT Media Lab, 2025

Literature review

Three principles for deciding how much effort an AI learning tool keeps.

1

AI makes tasks easier

Most AI tools follow these UX principles

  • Reduce memory load
  • Cut steps
  • Recognize, not recall

Learning science suggests some effort is needed to learn.

2

Learning takes effort

Confusion, trial, and reasoning come before understanding

Frictionless AI

Based on James Nottingham's Learning Pit and Seymour Papert's "hard fun"

3

The right setting depends on context

  • TaskFinish the task, or learn from doing it?
  • ExpertiseIs the learner ready for the effort, or do they need the steps first?
  • StakeDoes the learner care, and is it worth the effort?
  • CultureWho teaches, and which effort does this community value?

How we approached it

We combine learning science with speculative design:

Builds on James Nottingham's Learning Pit, Seymour Papert's "hard fun", and Julian Bleecker's design fiction.

  1. Map where AI removes effortCompare common AI UX principles with how learning happens.
  2. Prototype the futureA speculative workshop set in 2045 explores new forms of AI.

Beyond the chatbot

The workshop asks what embodied forms AI could take in learning.

Question Learner's effort Answer Elicit Learner tries first Restrict One hint at a time Reshape Forms of AI object ritual companion place ? TODAY Changes when the answer arrives 2045 Changes what form AI takes

Select Elicit, Restrict, or a form of AI.

Workshop

Method
Speculative design workshop
Premise
A time machine to a 2045 learning environment, adapted from Julian Bleecker
Task
Bring back an object, ritual, or practice that shows AI’s role in learning, seen through a different culture
Still from the workshop video: Come board the time machine Watch the workshop on Vimeo ↗

Where it stands

Ongoing research with Sneha Arvind. Get in touch to learn more.

  • ResultsHeld until publication.
  • NextTest the principles with learners and educators.