Intuit

Intuit: Explain Quantum Computing Concepts via AR-based Analogy

Background and motivation

Intuit is an augmented reality prototype for introducing quantum computing concepts through everyday objects and their virtual counterparts. The paper addresses the established educational difficulty of explaining superposition, measurement, entanglement, and related concepts to learners who lack the mathematical background needed for formal treatments. Its proposed contribution is a systematic way to select object analogies and implement them as interactive AR lessons, rather than the identification of a new learning problem.

The related work spans quantum education, quantum visualization, and situated analytics. Games such as quantum tic-tac-toe and Entanglion can make learning engaging, but still need guidance to connect their rules to quantum ideas. Hands-on quantum experiments and circuit programming offer experiential learning but may require equipment or expertise. The Bloch sphere, Q-sphere, and systems such as QuantumEyes and VIOLET support interpretation of states, circuits, or quantum neural networks, whereas Intuit focuses on the earlier task of building familiarity with basic concepts. Situated analytics supplies the design precedent for placing visual explanations in the physical environment and letting learners relate them to tangible objects. The authors argue that existing educational metaphors tend to be selected ad hoc and that unfamiliar quantum behavior is difficult to depict using physical objects alone. AR lets the system animate familiar objects in ways that would be impossible in the physical world, while maintaining a recognizable setting for the lesson.

Developing the analogy framework

The authors first reviewed textbooks, online courses, and research papers to identify fundamental concepts, then conducted a formative study with six quantum computing experts. These experts included two professors with more than five years of teaching experience and four Ph.D. students working in quantum computing. Two coauthors analyzed the literature and interview material, and domain experts validated the resulting concept list. The characterization framework emerged through four months of collaboration with experts, including two coauthors.

The framework maps four dimensions of a quantum concept to properties of an everyday object. The number of qubits maps to the number of objects used in the analogy. The authors' term “output state duality” distinguishes a representation that retains both basis-state possibilities from one that ends in a single outcome, and maps this distinction to rotation versus a stationary object. The concept type distinguishes states, processes, and operators, and maps to what the framework calls “translation.” Here translation means a change in the object's state, such as stopping its rotation, rather than only a change in spatial position. Finally, probability quantification maps to a continuously adjustable object property when the lesson explicitly works with probabilities, as in the gate demonstrations. The authors place lessons that emphasize qualitative behavior in the other category, including superposition and measurement; this is a classification of explanatory emphasis, not a claim that those quantum phenomena lack probabilities.

Figure 1 presents this framework through small tables and blue line drawings of candidate analogies. Coins, cards, and two-color spinners are alternatives for rotating and stopping objects; gears illustrate correlated motion; a paper cutter or a ruler-and-coin arrangement provides a continuous control for the gate examples. These are schematic proposals illustrating the mapping space. The implemented lessons shown in Figures 2 and 3 use coins, a blade-free paper cutter, and colored cubes, so the alternative sketches should not all be read as implemented interfaces.

Prototype, visual encoding, and interaction

Intuit consists of a Python server called IntuitVision and a Unity client called IntuitSense. The server performs object detection, including a YOLOv8-trained model for tracking the paper cutter and cubes, while the client coordinates inputs and the virtual scene. The prototype includes text definitions, mathematical expressions, auditory support, object manipulation, and bare-hand gestures. Making a fist triggers lesson actions, and a thumbs-up returns to the main menu. The implementation comprises seven modules and nine lessons: superposition, measurement, decoherence, tunneling, teleportation, entanglement, and separate lessons for Identity, Pauli-X, and Hadamard gates. Although the framework discusses gates involving up to three qubits, these implemented gate lessons demonstrate single-qubit gates.

In the superposition and measurement lessons, a learner places a physical coin in a highlighted circular region on the table. A fist gesture replaces its appearance with a rotating virtual coin, accompanied by a probability panel with labeled rows for $|0\rangle$ and $|1\rangle$. The illustrated superposition displays 50% in each row. Another fist gesture stops the rotation and shows a definite head or tail outcome, with the probability panel changing to 100% for the corresponding basis state. Figure 3A shows the sequence of real interactions and augmented views, together with enlarged callouts of the coin and probability panel. This explicitly connects a visible change in the object to a change in the displayed state probabilities.

The other coin lessons use animation to convey selected conceptual relationships. In Figure 2B, environmental effects accompany a slowing coin until it stops, serving as the paper's simplified analogy for decoherence. Figure 2C shows a rotating virtual coin passing through the table to illustrate tunneling, while Figure 2D depicts state transfer between coins at different locations for teleportation. Figure 3B depicts one entanglement scenario: measuring the left coin yields one basis-state outcome and simultaneously reveals the opposite outcome for the right coin. The paired probability panels reinforce that anticorrelation. These displays are pedagogical analogies; the paper does not present them as physical implementations or complete mathematical descriptions of the phenomena. For example, the two-coin teleportation animation conveys state transfer without implementing a complete teleportation protocol.

The gate lessons use slider position on a blade-free paper cutter to specify input probabilities. The bottom of the slider corresponds to $P(0)=0$, and the top to $P(0)=1$, with $P(0)+P(1)=1$ throughout. The paper sometimes refers to probability amplitudes when motivating the analogy, but its implemented slider description and Figure 3 explicitly show probabilities. A blue cube labeled “I,” a red cube labeled “X,” or a green cube labeled “H” selects the gate. A virtual paper cutter and probability panel show the output, and mathematical expressions explain the matrix operation. Figure 3C–E uses the same input $|1\rangle$ to make the transformations comparable: Identity preserves the input, Pauli-X changes it to $|0\rangle$, and Hadamard produces equal displayed probabilities. The equal probabilities in that example do not by themselves encode the full amplitude and phase information of a quantum state.

User evaluation and findings

The user study involved 16 university participants aged 21–33, including 15 Ph.D. students and one non-Ph.D. student. Their mean self-reported familiarity with quantum computing was 2.7 on a seven-point scale, and all had prior AR experience. After training on the object mappings and gestures, participants explored the nine lessons for 20 minutes, answered nine questions curated by a quantum computing expert, rated visualization effectiveness and usability, and discussed improvements. Sessions lasted 45–60 minutes. The study used a Quest 3 and an external camera for object tracking because the authors could not access the headset's camera data in their study setup.

Figure 4 combines a horizontal bar chart of correct quiz responses with stacked seven-point response distributions for perceived effectiveness and usability. Entanglement received 16 correct responses, superposition 15, Pauli-X 14, measurement, tunneling, and Identity 13 each, teleportation and decoherence 12 each, and Hadamard 7. Participants reported difficulty distinguishing the three gates and sometimes confused terminology such as tunneling and teleportation. The ratings were generally favorable, and comments described everyday objects as relatable and the animations as helpful. The plotted ratings for superposition, measurement, and entanglement are all at least 4/7; the prose describes them as above 4/7, but the chart also contains neutral responses at 4. Interaction smoothness received more mixed feedback, including two somewhat-disagree ratings associated with unreliable gesture recognition. One participant also reported delayed response in the Hadamard lesson.

These results support feasibility and favorable perceptions in this particular sample, with immediate post-session answers showing which concepts remained difficult. They do not establish a measured learning gain or superiority over another teaching method. The authors deliberately sought qualitative feedback rather than comparative performance and omitted a pre-test because participants reported little prior quantum knowledge. There was no comparison condition, and the paper reports no delayed retention assessment. The predominantly doctoral, AR-experienced sample also limits conclusions about broader novice populations.

Expert feedback, contributions, and future work

A separate evaluation interviewed six quantum computing experts with two to seven years of experience, four of whom had teaching experience. These experts had not participated in the formative study. They watched demonstration videos remotely and answered open-ended questions in 30–45-minute interviews; they did not interact with Intuit directly. They responded positively to combining analogies with AR, while identifying places where the representations needed more explanation. One expert suggested distinguishing the Hadamard output through two virtual paper cutters, and another recommended adding details such as the no-cloning theorem to clarify the teleportation lesson. This feedback supports the perceived educational promise of the approach, with a narrower evidential basis than hands-on expert usability testing.

The paper's main contributions are the expert-informed analogy characterization, its implementation in an AR teaching prototype, and exploratory evidence about how learners and experts respond to the lessons. It also exposes a practical tension: simplifying a concept through familiar motion can make it approachable, but the analogy still needs explicit explanations of what it represents and where its coverage ends. The results identify gates, especially Hadamard, as a priority for clearer visual distinctions and stronger connections between the objects and mathematical operations.

Participants requested adjustable animation speeds, more varied examples, and a progression from non-mathematical explanations to matrix derivations. The authors identify gesture-recognition reliability and testing with AR-naive learners as further needs. They propose extending the framework to quantum circuits and quantum neural networks, and adding adaptive learning features. These extensions are future directions rather than evaluated capabilities of the reported prototype.