Keystroke-level model: Difference between revisions

Revision as of 13:42, 3 September 2019

Revision as of 02:43, 19 December 2019

Line 1:

In [[human&ndash;computer interaction]], the '''keystroke-level model''' ('''KLM''') predicts how long it will take an expert user to accomplish a routine task without errors using an interactive computer system.<ref>{{cite journal|last1=Card|first1=Stuart K|last2=Moran|first2=Thomas P|last3=Allen|first3=Newell|title=The keystroke-level model for user performance time with interactive systems|journal=Communications of the ACM|date=1980|volume=23|issue=7|pages=396–410|doi=10.1145/358886.358895|ref=1}}</ref> It was proposed by [[Stuart K. Card]], [[Thomas P. Moran]] and [[Allen Newell]] in 1980 in the ''[[Communications of the ACM]]'' and published in their book ''The Psychology of Human-Computer Interaction'' in 1983, which is considered as a classic in the HCI field.<ref>{{cite web|last1=Sauro|first1=Jeff|title=5 Classic Usability Books|url=http://www.measuringu.com/blog/usability-books.php|website=MeasuringU|accessdate=22 June 2015|ref=7}}</ref><ref>{{cite web|last1=Perlman|first1=Gary|title=Suggested Readings in Human-Computer Interaction (HCI), User Interface (UI) Development, & Human Factors (HF)|url=http://hcibib.org/readings.html|website=HCI Bibliography : Human-Computer Interaction Resources|accessdate=22 June 2015|ref=8}}</ref> The foundations were laid in 1974, when Card and Moran joined the [[PARC (company)|Palo Alto Research Center]] (PARC) and created a group named Applied Information-Processing Psychology Project (AIP) with Newell as a consultant aiming to create an applied psychology of human-computer interaction.<ref>{{cite book|last1=Card|first1=Stuart K|last2=Moran|first2=Thomas P|last3=Newell|first3=Allen|title=The Psychology of Human-Computer Interaction|date=1983|publisher=L. Erlbaum Associates Inc|___location=Hillsdale|isbn=978-0898592436|pages=[https://archive.org/details/psychologyofhuma00stua/page/ ix–x]|ref=4|url=https://archive.org/details/psychologyofhuma00stua/page/}}</ref> The keystroke-level model is still relevant today, which is shown by the recent research about mobile phones and touchscreens (see [[#Adaptions|Adaptions]]).

== Structure of the keystroke-level model ==

Line 35:

|-

| K || total typing test time/total number of non-error keystrokes<br />

Guidelines:<ref name="klm-paper-operators-table">{{cite journal|last1=Card|first1=Stuart K|last2=Moran|first2=Thomas P|last3=Newell|first3=Allen|title=The keystroke-level model for user performance time with interactive systems|journal=Communications of the ACM|date=1980|volume=23|issue=7|pages=399|doi=10.1145/358886.358895|ref=5}}</ref><ref name="klm-book-operators-table">{{cite book|last1=Card|first1=Stuart K|last2=Moran|first2=Thomas P|last3=Newell|first3=Allen|title=The Psychology of Human-Computer Interaction|date=1983|publisher=L. Erlbaum Associates Inc|___location=Hillsdale|isbn=978-0898592436|pages=[https://archive.org/details/psychologyofhuma00stua/page/264 264]|ref=4|url=https://archive.org/details/psychologyofhuma00stua/page/264}}</ref><br />

.08 (135 wpm: best typist)<br />

.12 (90 wpm: good typist)<br />

Line 75:

== Comparison with GOMS ==

The KLM is based on the keystroke level, which belongs to the family of [[GOMS]] models.<ref>{{cite book|last1=Card|first1=Stuart K|last2=Moran|first2=Thomas P|last3=Newell|first3=Allen|title=The Psychology of Human-Computer Interaction|date=1983|publisher=L. Erlbaum Associates Inc|___location=Hillsdale|isbn=978-0898592436|pages=[https://archive.org/details/psychologyofhuma00stua/page/161 161–166]|ref=15|url=https://archive.org/details/psychologyofhuma00stua/page/161}}</ref> The KLM and the GOMS models have in common that they only predict behaviour of experts without errors, but in contrast the KLM needs a specified method to predict the time because it does not predict the method like GOMS.<ref>{{cite book|last1=Card|first1=Stuart K|last2=Moran|first2=Thomas P|last3=Newell|first3=Allen|title=The Psychology of Human-Computer Interaction|date=1983|publisher=L. Erlbaum Associates Inc|___location=Hillsdale|isbn=978-0898592436|pages=[https://archive.org/details/psychologyofhuma00stua/page/260 260]|ref=16|url=https://archive.org/details/psychologyofhuma00stua/page/260}}</ref> Therefore, the KLM has no goals and method selection rules, which in turn makes it easier to use.<ref name="klm-book-goms">{{cite book|last1=Card|first1=Stuart K|last2=Moran|first2=Thomas P|last3=Newell|first3=Allen|title=The Psychology of Human-Computer Interaction|date=1983|publisher=L. Erlbaum Associates Inc|___location=Hillsdale|isbn=978-0898592436|pages=[https://archive.org/details/psychologyofhuma00stua/page/269 269]|ref=17|url=https://archive.org/details/psychologyofhuma00stua/page/269}}</ref> The KLM resembles the model K1 from the family of GOMS models the most because both are at the keystroke level and possess a generic M operator. The difference is that the M operator of the KLM is more aggregated and thus larger (1.35 seconds vs. 0.62 seconds), which makes its mental operator more similar to the CHOOSE operations of the model K2.<ref name="klm-book-goms" /> All in all, the KLM represents the practical use of the GOMS keystroke level.<ref>{{cite book|last1=Card|first1=Stuart K|last2=Moran|first2=Thomas P|last3=Newell|first3=Allen|title=The Psychology of Human-Computer Interaction|date=1983|publisher=L. Erlbaum Associates Inc|___location=Hillsdale|isbn=978-0898592436|pages=[https://archive.org/details/psychologyofhuma00stua/page/264 264]|ref=18|url=https://archive.org/details/psychologyofhuma00stua/page/264}}</ref>

== Advantages ==

Line 82:

== Limitations ==

The keystroke-level model has several restrictions:

* It measures only one aspect of performance: time,<ref>{{cite journal|last1=Card|first1=Stuart K|last2=Moran|first2=Thomas P|last3=Newell|first3=Allen|title=The keystroke-level model for user performance time with interactive systems|journal=Communications of the ACM|date=1980|volume=23|issue=7|pages=400|doi=10.1145/358886.358895|ref=21}}</ref> which means execution time and not the time to acquire or learn a task <ref>{{cite book|last1=Card|first1=Stuart K|last2=Moran|first2=Thomas P|last3=Newell|first3=Allen|title=The Psychology of Human-Computer Interaction|date=1983|publisher=L. Erlbaum Associates Inc|___location=Hillsdale|isbn=978-0898592436|pages=[https://archive.org/details/psychologyofhuma00stua/page/260 260–261]|ref=22|url=https://archive.org/details/psychologyofhuma00stua/page/260}}</ref>

* It considers only expert users. Generally, users differ regarding their knowledge and experience of different systems and tasks, motor skills and technical ability <ref name="klm-paper-limitations-users">{{cite journal|last1=Card|first1=Stuart K|last2=Moran|first2=Thomas P|last3=Newell|first3=Allen|title=The keystroke-level model for user performance time with interactive systems|journal=Communications of the ACM|date=1980|volume=23|issue=7|pages=397, 409|doi=10.1145/358886.358895}}</ref>

* It considers only routine unit tasks <ref name="klm-paper-limitations">{{cite journal|last1=Card|first1=Stuart K|last2=Moran|first2=Thomas P|last3=Newell|first3=Allen|title=The keystroke-level model for user performance time with interactive systems|journal=Communications of the ACM|date=1980|volume=23|issue=7|pages=409|doi=10.1145/358886.358895|ref=24}}</ref>

* The method has to be specified step by step<ref name="klm-paper-limitations" />

* The execution of the method has to be error-free <ref name="klm-paper-limitations" />

* The mental operator aggregates different mental operations and therefore cannot model a deeper representation of the user’s mental operations. If this is crucial, a GOMS model has to be used (e.g. model K2)<ref>{{cite book|last1=Card|first1=Stuart K|last2=Moran|first2=Thomas P|last3=Newell|first3=Allen|title=The Psychology of Human-Computer Interaction|date=1983|publisher=L. Erlbaum Associates Inc|___location=Hillsdale|isbn=978-0898592436|pages=[https://archive.org/details/psychologyofhuma00stua/page/285 285–286]|ref=25|url=https://archive.org/details/psychologyofhuma00stua/page/285}}</ref>

Also, one should keep in mind when assessing a computer system that other aspects of performance (errors, learning, functionality, recall, concentration, fatigue, and acceptability),<ref>{{cite journal|last1=Card|first1=Stuart K|last2=Moran|first2=Thomas P|last3=Newell|first3=Allen|title=The keystroke-level model for user performance time with interactive systems|journal=Communications of the ACM|date=1980|volume=23|issue=7|pages=396–397|doi=10.1145/358886.358895|ref=26}}</ref> types of users (novice, casual)<ref name="klm-paper-limitations-users" /> and non-routine tasks have to be considered as well.<ref name="klm-paper-limitations-users"/>

Furthermore, tasks which take more than a few minutes take several hours to model and a source of errors is forgetting operations.<ref>{{cite book|last1=Sauro|first1=Jeff|editor1-last=Jacko|editor1-first=Julie A|title=Estimating productivity: Composite operators for keystroke level modeling|journal=Human-Computer Interaction. New Trends: Proceedings of the 13th International Conference (LNCS)|date=2009|volume=5610|page=353|doi=10.1007/978-3-642-02574-7_40|ref=28|publisher=Springer-Verlag|___location=Berlin Heidelberg|series=Lecture Notes in Computer Science|isbn=978-3-642-02573-0}}</ref> This implies that the KLM is best suited for short tasks with few operators.

In addition, the KLM can not make a perfect prediction and has a root-mean-square error of 21%.<ref>{{cite book|last1=Card|first1=Stuart K|last2=Moran|first2=Thomas P|last3=Newell|first3=Allen|title=The Psychology of Human-Computer Interaction|date=1983|publisher=L. Erlbaum Associates Inc|___location=Hillsdale|isbn=978-0898592436|pages=[https://archive.org/details/psychologyofhuma00stua/page/275 275]|ref=29|url=https://archive.org/details/psychologyofhuma00stua/page/275}}</ref>

== Example ==

Line 132:

== Adaptions ==

The six operators of the KLM can be reduced, but this decreases the accuracy of the model. If this low of an accuracy makes sense (e.g. “back-of-the-envelope” calculations) such a simplification can be sufficient.<ref>{{cite book|last1=Card|first1=Stuart K|last2=Moran|first2=Thomas P|last3=Newell|first3=Allen|title=The Psychology of Human-Computer Interaction|date=1983|publisher=L. Erlbaum Associates Inc|___location=Hillsdale|isbn=978-0898592436|pages=[https://archive.org/details/psychologyofhuma00stua/page/296 296]|ref=34|url=https://archive.org/details/psychologyofhuma00stua/page/296}}</ref>

While the existing KLM applies to desktop applications, the model might not fulfill the range of mobile tasks,<ref>{{cite book|last1=Li|first1=Hui|last2=Liu|first2=Ying|last3=Liu|first3=Jun|last4=Wang|first4=Xia|last5=Li|first5=Yujiang|last6=Rau|first6=Pei-Luen Patrick|title=Extended KLM for mobile phone interaction: a user study result|journal=CHI EA '10 CHI '10 Extended Abstracts on Human Factors in Computing Systems|date=2010|publisher=ACM|___location=New York|isbn=978-1-60558-930-5}}</ref> or as Dunlop and Cross <ref>{{cite journal|last1=Dunlop|first1=M.|last2=Crossan|first2=A.|title=Predictive Text Entry Methods for Mobile Phones|journal=Personal Technologies|date=2000|pages=134–143}}</ref> declaimed KLM is no longer precise for mobile devices. There are various efforts to extend the KLM regarding the use for mobile phones or touch devices. One of the significant contributions to this field is done by Holleis, who retained existing operators while revisiting the timing specifications. Furthermore, he introduced new operators: Distraction (X), Gesture (G), Initial Act (I).