Keystroke level model

The Keystroke Level Model (KLM) allows statements to be made about how long an expert needs to process a task with an interactive computer system without errors. In the research area of human-computer interaction, the KLM is one of the most widespread approaches and is mainly applied to word processing tasks. In order to predict the required working time, a sequence of interaction steps, a so-called method, is specified. The execution time then results from the sum of the operators of the individual work steps. By adapting the model to mobile devices and touchscreens, the KLM is still relevant today.

Structure of the keystroke level model

The keystroke level model comprises six operators, of which the first four operators are defined as physical-motor operators. There is also a mental operator and an operator that represents the response time of the system.

K (keystroke or mouse click): This is the most common operator. However, it does not only describe letters, but every single keystroke (e.g. a keystroke on the shift key (SHIFT) is evaluated as a single K operation). The time required depends on the motor skills of the user and is determined by means of one-minute typing tests, the test time being divided by the total number of error-free keystrokes.
P (target acquisition on a display with the help of a mouse): This time depends on the distance to the target and the size of the target. A mouse click is not included. It counts as a separate K operation.
H (change of hand or hands between different input devices): The time for this operator includes the movement when changing input devices or the fine positioning of the hand.
D (drawing (manual): n _D straight line segments with a total length of D (n _D , l _D ) cm): where n _{D represents} the number of line segments and l _D the total length of the line segments. This operator is very restrictive , because it is assumed that a mouse is used for drawing, with the drawing system limiting the cursor to a 0.56 cm rectangle. The calculated time for this operator is an average.
M (mental preparation to perform physical actions): Describes the time a user needs for thought processes or decision making. The number of M-operators in a method depends on the knowledge and skills of the user. Heuristics provide information about where in a method an M operator must be set. If, for example, a target is controlled with the mouse, a mouse click is usually also required, which is why no M operator is required between these two operators. Point 2 explains the heuristics for setting an M-operator.
R (response time of the system): The response time depends on the system, the command and the context of the command. It is only needed when the user actually has to wait for the system. For example, if the user prepares his next physical action (M), only the non-overlapping part of the response time is required for R, since the user needs the response time for the M operation (for example R of 2 seconds - M for 1.35 seconds = R of .65 seconds). Kieras suggests changing the name of the operator to W (wait) to avoid confusion. Sauro, on the other hand, prefers a test that determines the system's response time in advance.

In addition to an overview of the times for the operators mentioned, the following table also shows times for other operators proposed by different authors.

operator	Time (seconds)
K	Total time of writing test divided by total number of error-free keystrokes Guidelines: .08 (135 wpm: best typist) .12 (90 wpm: good typist) .20 (55 wpm: average typist) .28 (40 wpm: average typist (not secretary)) .50 (typing random letters) . 75 (writing a complex code) 1.20 (worst typist and not familiar with the keyboard)
P	1.1
H	0.4
D.	.9n _D +. 16 l _D
M.	1.35
R.	System dependent
Other suggested operators
B (mouse click or release of the mouse button)	0.1
Click on a button / link	3.73
Pull-down list (without loading the page)	3.04
Pull-down list (when the page is loaded)	3.96
Date picker	6.81
Cut & Paste (keyboard)	4.51
Write text in a text field	2.32
Scrolling	3.96

Methodology for setting the M operator

The procedure for setting the M operator is as follows:

Start with a method that includes all physical operators and answer operations. Use rule 0 to place an M-operator. For each M, iterate through rules 1 through 4 to see if it should be removed.
Rule 0	Put Ms in front of all K operators that are not part of a continuous character string (for example, texts or numbers). Put Ms in front of all P-operators that select commands (no arguments).
Rule 1	If an operator after an M can already be anticipated in the operator before the M, then remove the M (for example PMK -> PK).
Rule 2	If a chain of MKs belongs to a cognitive unit, delete all Ms except for the first one.
Rule 3	If a K is a redundant terminator (e.g. the end of a command immediately after the end of an argument), then remove the M in front of it.
Rule 4	If a K ends a constant string (e.g. a command name), remove the M in front of it. However, if a K ends a variable character string (for example an argument), the M must remain.

Compared to GOMS

The KLM is the simplest version of the GOMS technology. One of the main differences between the KLM and other models in the GOMS family is that the KLM does not predict what a user will do, but rather specifies them. Therefore, in contrast to the GOMS model, the KLM does not contain any goals or selection rules. In addition, unlike other GOMS models, the KLM is limited to tasks in which there are no parallel work steps, nested goals or interruptions.

advantages

The KLM is a quick and easy model that can be used without deeper knowledge of human psychology. In addition, turnaround times can be predicted without prototyping and testing with actual users, saving time and money.

restrictions

The KLM is limited by several factors:

Only time is measured and other performance measures are not recorded.
It can only predict times for experts. In general, users differ in their previous knowledge and experience with different systems and tasks, through motor skills and technical skills.
Only known and routine tasks are considered.
The method must be specified step by step.
The execution of the method must be error-free.
The M-operator comprises different mental operations and does not give any more precise information about the mental operations of the user. If this is necessary, a GOMS model must be used.

In general, one should be aware that when dealing with computer systems, other performance measures (errors, learning, functionality, recall, concentration, fatigue and task appropriateness) as well as other types of users (beginners, occasional users) and non-routine tasks also play a role. In addition, it usually takes several hours to make a forecast for tasks that take longer than a few minutes. The fact that operations are forgotten is also a frequent source of error. This means that the KLM is best suited for short tasks with a small number of operators. In addition, the KLM cannot make a really precise prediction and, with a mean square deviation of 21%, is often incorrect.

example

The following example, an adaptation of a calculation by Kieras, shows the practical use of the KLM. Here, two possibilities of deleting a file are compared, whereby the person to be executed is an average typist. For the M-Operator, 1.35 seconds, as specified in the KLM, are used instead of the 1.2 seconds used by Kieras. The time difference remains unchanged in this case.

Design A: Drag the file to the trash	Design B: Using the shortcut “control + T”
Encoding methods: command chain	Encoding methods: command chain
Prepare to delete (M) Finding the file icon (M) Navigate to the file icon (P) Press and hold the mouse button (B) Drag the file to the trash (P) Release the mouse button (B) Point to the origin window (P)	Prepare to delete (M) Finding the file icon (M) Navigate to the file icon (P) Press and hold the mouse button (B) Release the mouse button (B) Move hand to keyboard (H) Press the control button (K) Press the T button (K) Move hand to mouse (H)
Needed time	Needed time
3P + 2B + 2M = 3 * 1.1 sec + 2 * .1 sec + 2 * 1.35 sec = 6.2 seconds	P + 2B + 2H + 2K + 2M = 1.1 sec + 2 * .1 sec + 2 * .4 sec + 2 * .2 sec + 2 * 1.35 sec = 5.2 seconds

This shows that Design B is faster even though it involves more operations.

Adaptation

The six operators of the KLM can be further reduced, but this reduces the precision of the model. However, this is quite practical for rough calculations.

Since the existing KLM is adapted to desktop computers, it only applies to a limited extent to mobile devices. Dunlop and Cross point out that the KLM is no longer correct for the mobile context. There are some attempts to expand the model so that it can also be used for mobile devices and touch surfaces.

A significant contribution to this research question comes from Holleis et al., Who keep the existing operators and only change the time spans of them. They also introduce other new operators: distraction (X), gesture (G) and initial action (IA). While Holleis et al. and Li et al. agree that the KLM can also be used for mobile devices, suggest Li et al. proposed another adaptation of the model. So they introduce a new concept, the operator blocks. This is "a frequently recurring sequence of operators of the extended model". In addition, Li et al. the original operators and define five new mental and nine new physical operators, with four of the physical operators being pen-based.

Rice and Lartigue designed the touch level model. To do this, they not only adapt the existing operators, but also suggest several operators for touch devices. They keep the operators Keystroke (K), Homing (H), Mental Act (M) and Response Time (R (t)) and suggest new ones based on the calculations by Holleis et al. based touch-specific operators:

Distraction: A multiplicative operator that adds time to other operators.
Pinch: A gesture to zoom out made with two fingers or more.
Zoom: A gesture to zoom in made with two fingers or more.
Initial Act: The steps that are required to prepare the system for use (for example, unlocking the device, pressing an icon, entering a password).
Tap: Tap the screen to start an action or make a change.
Swipe: A gesture in which one or more fingers are placed on the screen and then repeatedly moved in one direction for a specified amount of time.
Tilt: Tilt or rotate the end device by d degrees.
Rotate: A gesture in which two or more fingers are placed on the screen and then rotated d degrees around an axis.
Drag: A gesture of placing one or more fingers on the screen and then moving them to another point, often in a straight line.

Web links

GOMS article by Lorin Hochstein
KLM exercise by Christoph Erhardt
Simple KLM calculator (free, downloadable Windows app)

Individual evidence

↑ ^a ^b ^c ^d ^e ^f ^g ^h ⁱ ^j ^k ^l ^m ⁿ ^o ^p ^q ^r Stuart K. Card, Thomas P. Moran, Newell Allen: The keystroke-level model for user performance time with interactive systems . In: Communications of the ACM . Vol. 23, No. 7 , 1980, doi : 10.1145 / 358886.358895 .
↑ Peter Haunold: Analysis of manual digitization processes with the keystroke level model . In: Applied geographic information processing . Vol. 26, 1997 ( PDF ). PDF ( Memento of the original from March 4, 2016 in the Internet Archive ) Info: The archive link was inserted automatically and has not yet been checked. Please check the original and archive link according to the instructions and then remove this notice. @1@ 2
^ ^A ^b Bonnie E. John, David E. Kieras: The GOMS Family of User Interface Analysis Techniques: Comparison and Contrast . In: ACM Transactions on Computer-Human Interaction (TOCHI) . Vol. 3, No. 4 , 1996, doi : 10.1145 / 358886.358895 .
^ Paul M. Fitts: The information capacity of the human motor system in controlling the amplitude of movement . In: Journal of Experimental Psychology: General . Vol. 121, No. 3 , 1992, doi : 10.1037 / h0055392 .
↑ ^a ^b ^c ^d ^e ^f David Kieras: Using the Keystroke-Level Model to Estimate Execution Times .
↑ ^a ^b ^c ^d ^e ^f ^g ^h ⁱ ^j Jeff Sauro, Julie A. Jacko (Editor): Estimating productivity: Composite operators for keystroke level modeling . In: Human-Computer Interaction. New Trends: Proceedings of the 13th International Conference (LNCS) . Vol. 5610. Springer-Verlag, Berlin Heidelberg 2009, doi : 10.1007 / 978-3-642-02574-7_40 .
↑ ^a ^b ^c ^d ^e ^f ^g ^h ⁱ ^j ^k Stuart K. Card, Thomas P. Moran, Allen Newell: The Psychology of Human-Computer Interaction . L. Erlbaum Associates Inc, Hillsdale 1983, ISBN 0-89859-243-7 .
↑ ^a ^b Hui Li, Ying Liu, Jun Liu, Xia Wang, Yujiang Li, Pei-Luen Patrick Rau: Extended KLM for mobile phone interaction: a user study result . In: CHI EA '10 CHI '10 Extended Abstracts on Human Factors in Computing Systems . ACM, New York 2010, ISBN 978-1-60558-930-5 , doi : 10.1145 / 1753846.1754011 .
↑ M.Dunlop, A. Crossan: Predictive Text Entry Methods for Mobile Phones . In: Personal Technologies . 2000, doi : 10.1007 / BF01324120 .
↑ ^a ^b ^c P. Holleis, F. Otto, H. Hussmann, A. Schmidt: Keystroke-level model for advanced mobile phone interaction . In: CHI '07: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems . 2007, doi : 10.1145 / 1240624.1240851 .
^ AD Rice, JW Lartigue: Touch-Level Model (TLM): Evolving KLM-GOMS for Touchscreen and Mobile Devices Categories and Subject Descriptors . In: ACM SE '14 Proceedings of the 2014 ACM Southeast Regional Conference Article No. 53 . 2014, doi : 10.1145 / 2638404.2638532 .

[klm-original-1] ↑ ^a ^b ^c ^d ^e ^f ^g ^h ⁱ ^j ^k ^l ^m ⁿ ^o ^p ^q ^r Stuart K. Card, Thomas P. Moran, Newell Allen: The keystroke-level model for user performance time with interactive systems . In: Communications of the ACM . Vol. 23, No. 7 , 1980, doi : 10.1145 / 358886.358895 .

[2] Peter Haunold: Analysis of manual digitization processes with the keystroke level model . In: Applied geographic information processing . Vol. 26, 1997 ( PDF ). PDF ( Memento of the original from March 4, 2016 in the Internet Archive ) Info: The archive link was inserted automatically and has not yet been checked. Please check the original and archive link according to the instructions and then remove this notice. @1@ 2

[goms-comparison-3] A ^b Bonnie E. John, David E. Kieras: The GOMS Family of User Interface Analysis Techniques: Comparison and Contrast . In: ACM Transactions on Computer-Human Interaction (TOCHI) . Vol. 3, No. 4 , 1996, doi : 10.1145 / 358886.358895 .

[4] Paul M. Fitts: The information capacity of the human motor system in controlling the amplitude of movement . In: Journal of Experimental Psychology: General . Vol. 121, No. 3 , 1992, doi : 10.1037 / h0055392 .

[estimate-execution-times-5] ↑ ^a ^b ^c ^d ^e ^f David Kieras: Using the Keystroke-Level Model to Estimate Execution Times .

[sauro-suggested-operators-6] ↑ ^a ^b ^c ^d ^e ^f ^g ^h ⁱ ^j Jeff Sauro, Julie A. Jacko (Editor): Estimating productivity: Composite operators for keystroke level modeling . In: Human-Computer Interaction. New Trends: Proceedings of the 13th International Conference (LNCS) . Vol. 5610. Springer-Verlag, Berlin Heidelberg 2009, doi : 10.1007 / 978-3-642-02574-7_40 .

[klm-erweitert-7] ↑ ^a ^b ^c ^d ^e ^f ^g ^h ⁱ ^j ^k Stuart K. Card, Thomas P. Moran, Allen Newell: The Psychology of Human-Computer Interaction . L. Erlbaum Associates Inc, Hillsdale 1983, ISBN 0-89859-243-7 .

[li_et_al.-8] Hui Li, Ying Liu, Jun Liu, Xia Wang, Yujiang Li, Pei-Luen Patrick Rau: Extended KLM for mobile phone interaction: a user study result . In: CHI EA '10 CHI '10 Extended Abstracts on Human Factors in Computing Systems . ACM, New York 2010, ISBN 978-1-60558-930-5 , doi : 10.1145 / 1753846.1754011 .

[9] M.Dunlop, A. Crossan: Predictive Text Entry Methods for Mobile Phones . In: Personal Technologies . 2000, doi : 10.1007 / BF01324120 .

[holleis_et_al.-10] P. Holleis, F. Otto, H. Hussmann, A. Schmidt: Keystroke-level model for advanced mobile phone interaction . In: CHI '07: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems . 2007, doi : 10.1145 / 1240624.1240851 .

[11] AD Rice, JW Lartigue: Touch-Level Model (TLM): Evolving KLM-GOMS for Touchscreen and Mobile Devices Categories and Subject Descriptors . In: ACM SE '14 Proceedings of the 2014 ACM Southeast Regional Conference Article No. 53 . 2014, doi : 10.1145 / 2638404.2638532 .