Fitts' Law as a Research and Design Tool in Human-Computer Interaction

Author: I. Scott MacKenzie

thesis

A critical assessment of Fitts’ law and suggested adaptations for research and design

supporting arguments

1. It is difficult to measure human behavior with precise models, as it is highly variable (^4f2492)
2. Corrections
   1. Welford’s variation is most common; it is similar to Shannon’s information theory, which Fitts’ law was derived from
   2. Shannon formulation
   3. Adjusting width based on subject performance
3. Connect the theory to practical problems in HCI by surveying six studies applying Fitts’ law
   1. Negative intercepts → Fitts’ law is good for more complicated tasks (higher ID)
   2. Lack of consensus in the measures
4. A good model to shape thinking, but not entirely accurate
5. Still needs more research before being integrated into design tools

methods and theory

1. Models “facilitate the way we think about a problem” (114), hence there is no right or wrong
2. Principles
   1. Two dimensions: shape of target and direction of movement
3. Common errors
   1. Undershoot errors
4. Strengths 2. Affords physical interpretation 3. Performs well for various limb and muscle groups
5. Weaknesses
   1. The current model is information focused, but still awaits verification through a theory of human movement (100)
   2. Failure of the model when ID is small
      1. Can avoid negative task difficulty by
         1. Shannon formulation of Fitts’ law
         2. Adjustment of target width based on what subject does
   3. Many sources of variation
   4. No standardized task for measurement
   5. No sound technique for error handling
   6. Inconsistencies in Fitts’ law studies’ predictions
6. Considerations
   1. Learning effects
   2. Range of conditions

terms and themes

• Index of performance (IP)
• Index of difficulty (ID)
• Movement time (MT)
• Fitts’ law equation: where a and b are regression coefficients $IP=ID/MT$ $ID=lo{g}_2(2A/W)$ $MT=a+blo{g}_2(2A/W)$
• Human channel: information transmission
• Human channel capacity: not straightforward due to random error
• Target width (W)
• Target amplitude (A) i.e. distance travelled
• Shannon formulation (most commonly used now)

$$MT=a+b{\log }_2(A/W+1)$$

questions

• How can the model be extended to varying approach angles and target shapes?
• How can Fitts’ law be improved such that it can be integrated into research and design tools?

notes

“Certainly though, conducting empirical experiments to validate models is the starting point. Putting the theory into tools comes later. When properly applied and integrated into tools, however, theories may indeed elicit new ways of thinking for designers” (94) “Fitts’ law may be placed in this continuum. As a mathematical expression, it emerged from the rigors of probability theory, yet when applied to psychomotor behavior it becomes a metaphor.” (98–99)

conclusion

“These issues plus a proper understanding of device-task associations play a vital role int eh development of Fitts’ law performance models capable of participating in the design of natural, efficient human-computer interfaces” (134)