Quantifying Aversion to Costly Typing Errors in Expert Mobile Text Entry

How do People Type on Mobile Devices?

“…Recent work found that this may cause adjustments in a speed-accuracy trade-off. Users may for example slow down to minimize the risk of errors [5]. Cognitive and motor impairments, such as dyslexia, tremor, or memory dysfunction, and various effects of aging, can have a detrimental effect on typing performance.…”

“…Accordingly, the amount of backspacing is also lower (1.89 versus 2.29 on average). A possible explanation is the higher interaction cost of correcting mistakes on mobile devices and the limited text editing methods (see [5] for a discussion). Nevertheless, KSPC is remarkably similar (M = 1.17, SD = 0.09 in [11]) with only the standard deviation being smaller compared with desktop keyboard entry potentially due to the varying use of intelligent text entry methods on mobile devices (see below).…”

“…This paper contributes to efforts in understanding typing performance with mobile devices, a central topic in recent HCI research (e.g. [3,5,7,29,33,35]). Mobile devices are extensively used for text input, in activities such as email, internet browsing, texting, and social media [9].…”

A glimpse of mobile text entry errors and corrective behaviour in the wild

“…So far we have observed that our participants type generally slowly and make few unnoticed mistakes at the word level (though every session contains two such mistakes on average). In [4], it is shown that expert typists will generally slow down to avoid the cost of correcting entry mistakes, increasing their movement time. In this cost, motor performance is just one part of the equation: the error has to be first detected, hence possibly requiring a frequent context-switch between the entered text and the keyboard (this is also acknowledged in [4]).…”

“…In [4], it is shown that expert typists will generally slow down to avoid the cost of correcting entry mistakes, increasing their movement time. In this cost, motor performance is just one part of the equation: the error has to be first detected, hence possibly requiring a frequent context-switch between the entered text and the keyboard (this is also acknowledged in [4]). In [11], it is shown that typists have a "stopping" span, i.e.…”