Author: Mingrui “Ray” Zhang
purpose
This paper establishes text-entry throughput as a single unified measure of input efficiency due to its stability across different speed-accuracy biases
subjects and sites
- Between-subjects
- 27 participants
- 15 laptop keyboard
- 12 smartphone keyboard
- Experienced typists aged 22–27
methods and theory
- Calculate source information
- Calculate transmission probabilities
- Calculate throughput
terms and themes
- Speed-accuracy bias: subjective and internal, can change with context, affects performance
- Shannon’s information theory
- Presented string (P)
- Transcribed string (T)
questions
- “How can we draw firm performance conclusions in the presence of [speed-accuracy] tradeoffs?” (1)
- How do we “reconcile speed with uncorrected errors?” (1)
- “Can we successfully manipulate different speed-accuracy biases for participants during text entry?” (7)
- “How stable is our new throughput metric across different participant speed-accuracy biases?” (7)
- “How does throughput stability compare to that of established speed and error rate metrics?” (7)
notes
“there was no significant increase in throughput using auto-correction, word completion, and word prediction” (12) “Our hope is that throughput will be calculated and reported to support comparisons across devices, text entry methods, and participants” (12)
- Speed and accuracy should also be reported due to their practical insight