ExNovo: A user interface that supports novices without encumbering experts
Mark Blair, Amanda Klassen, Justin O'Camb, Cal Woodruff, Rollin Poe, Christine Chuong, Robin Barrett · 2026 · ACM Games · doi:10.1145/3786596
Summary
ExNovo is a novel hierarchical tree menu interface designed to bridge the long-standing divide between graphical user interfaces (GUIs) and keyboard hotkeys. Traditional GUIs scaffold novice users by making command options visible, but their reliance on precise mouse targeting prevents users from transitioning to automatic, expert-level performance. Hotkeys enable expert speed by requiring only stable finger movements, but demand prior memorization of arbitrary key combinations, making them hard to learn. ExNovo unifies these approaches through a hierarchical ternary-branching tree: commands are selected via short sequences of three physical keys (numbered 1, 2, 3), and each node in the visual menu is labeled with the sequence required to select it. This means using the GUI as a novice simultaneously rehearses the expert motor sequences. The interface is navigated with one hand using only five keys (three selection keys, one Confirm, one Cancel), eliminating hand repositioning. Commands are organized into 27 "neighborhoods" of conceptually related actions, supporting memorization through chunking. Colors and auditory tones provide additional multi-sensory encoding. The paper reports three experiments. The first two tested ExNovo against a standard right-click contextual menu in a Space Invaders-style game (193 and 32 participants respectively). The third scaled the interface to approximately 400 commands with sequences up to six keypresses, training four participants across 20 sessions over several months. The work addresses how interface design can simultaneously support users across the full novice-to-expert spectrum, and discusses implications for accessibility, games, and extended reality environments.
Key findings
Experiment 1 found that right-click menu users outperformed ExNovo users during the initial stream phase (1,184.9ms faster time-to-kill, 17.94% higher accuracy), but this advantage vanished by the shootout phase after less than 30 minutes of practice, with no significant differences in speed or accuracy. Experiment 2 extended training to two sessions one week apart; ExNovo users were significantly faster than right-click users in the shootout phases (β = −421.87ms, p = 0.014), demonstrating that ExNovo surpasses a standard GUI with modest additional practice. Experiment 3 showed that scaling to ~400 commands caused performance drops at each complexity transition but that performance continued to improve with practice and never plateaued. By session 20, participants achieved ~15 hits per minute, recovering roughly half the performance lost from scaling up. Median sequence latency was approximately 1–2 seconds for short sequences and 3–5 seconds for length-6 sequences; 90% of all responses fell under 10 seconds even for the longest sequences. Structural properties of the command tree significantly predicted performance: sequence length (β = 879.6ms per additional keystroke), sequence randomness (β = 1,114.6ms), and number of next-door neighbors (β = −267.3ms, with nodes having more neighbors being faster to select). A user survey found participants rated hotkeys as efficient but used only about four regularly, reflecting the well-known failure to transition from GUI to expert use.
Relevance
ExNovo has direct implications for digital accessibility. Its one-handed five-key design reduces motor demands compared to point-and-click interfaces, removing the need for precise cursor targeting — a barrier for users with motor impairments. Because menu selections are read with only a few options per level and forward progress is maintained, the interface is naturally compatible with screen readers and voice guidance, making it relevant for users with visual impairments. The auditory tones encoding sequences provide non-visual feedback. The authors note that ExNovo's restricted input set coordinates well with a variety of assistive devices. Its applicability to extended reality (XR) environments — where traditional GUI navigation is awkward — suggests value for emerging AT platforms. The paper's analysis of the Paradox of the Active User helps explain why hotkey training approaches fail, and its core design insight — that users should learn expert behaviors incidentally through normal task performance — is a principle directly applicable to assistive technology interface design and AT adoption barriers.
Tags: user interface design · keyboard accessibility · motor accessibility · visual impairment · novice-to-expert transition · human-computer interaction · one-handed input · cognitive load · hotkeys · game interfaces · extended reality