Maximizing the guessability of symbolic input

TLDR

Guessability is essential for symbolic input, where users rely on gestures, keywords, labels, or icons to indicate characters or commands. The authors propose a unified method for maximizing and evaluating symbolic input guessability, offering formulae for quantification and agreement that designers and evaluators can use. The method computes guessability metrics and agreement, then tests symbol sets using the MacKenzie and Zhang (1997) usability procedure. Applying the method to the EdgeWrite unistroke alphabet raised guessability from 51.0 % to 80.1 % and improved learning accuracy from 78.8 %/90.2 % to 81.6 %/94.2 %, approaching Graffiti’s 81.8 %/95.8 % performance.

Abstract

Guessability is essential for symbolic input, in which users enter gestures or keywords to indicate characters or commands, or rely on labels or icons to access features. We present a unified approach to both maximizing and evaluating the guessability of symbolic input. This approach can be used by anyone wishing to design a symbol set with high guessability, or to evaluate the guessability of an existing symbol set. We also present formulae for quantifying guessability and agreement among guesses. An example is offered in which the guessability of the EdgeWrite unistroke alphabet was improved by users from 51.0% to 80.1% without designer intervention. The original and improved alphabets were then tested for their immediate usability with the procedure used by MacKenzie and Zhang (1997). Users entered the original alphabet with 78.8% and 90.2% accuracy after 1 and 5 minutes of learning, respectively. The improved alphabet bettered this to 81.6% and 94.2%. These improved results were competitive with prior results for Graffiti, which were 81.8% and 95.8% for the same measures.