HCI( Human Computer Interaction) Fitts Fitts ISO HCI Fitts IC Fitts SH Systematic Human-factor Model for Pointing Task Evaluation Xiangsi REN, Xing-Qi JING Koci University of Tecnology, saikawa University One callenging job for tose wo are doing researc on Fitts law, an evaluation tool used in uman computer interaction, is to resolve te problem of te input its distribution. ltoug one metod as been proposed and used for many years as te ISO standard for one -direction pointing task, doubts now exist regarding its validity. Design tasks in HCI demand a revision or a replacement of te model offered in Fitts law. In response to tis need, we developed a new model for evaluating input efficiency of pointing tasks. Te prediction effect of te new and te traditional models are compared using IC (kaike Information Criterion), a criterion for statistical model selection. Te results of IC evaluation sow tat te new model, developed by us and described in tis paper, is better tan te traditional ones in performance prediction and evaluation. Fitts 954 Fitts [5] Fitts [,0] ccot [] Fitts Fitts
(IP:index of performance) 4% Sannon (cannel Fitts capacity) [4] IP [9] e (effective target widt) IP= ID/ MT () ISO 94-9 [3] 4% MT ID (index of difficulty) ISO 94-9 Fitts [6] ID= log ( / ) () Fitts (add oc) Fitts (post oc) 4% () MT = a + bid (3) Fitts (3) a b MT (3) ID Fitts Fitts Sannon 7 Sannon [8] MT = a + b log ( + ) (4) + Ps Ps = (6) ID = log ( + ) (5) + (4) (5) (6) Ps = (7) Sannon + ( λ + ) (GN: dditive ite Gaussian Noise) λ [4] (self-information) [9,0] ID = s log log ( ) P = s + λ + (8) 96% 96% 4% Sannon 7 C e C=Blog (S/N+) B MT = a + blog ( + ) S/N [4] e
+ ( λ + ) + (8) ID (5) IDs s ID λ ID IC () () SH SH-Model: S H Systematic P Human ID s λ P + = ID P P (self-information) ID maximum likeliood ID = log ( ) (9) estimation P P n P = (3) m P n m MT = a + bid + cid (0) s (3)P MT a b c P = 0 (0) P (uniform prior [5] distribution) P [] n + P = (4) m + [4,7,] 3 ln( MT) = a' + b'ln( ID ) + c'ln( ID ) () s a b c ab c ()ln( ID )ln( ID ) s ln(mt) MT real k ln( MT ) = a' + b'ln( ID ) + c'ln( ID ) + k () real s P = P () 3
IC IC Sannon MT = 97.39+ 75.3log ( + ) 3897 e MT = 5.05+ 65.3log ( + ) 39078 e SH MT = e log ( + ) log ( ) 37696 5.734 0.6365-0.0755 P MT = e ID ID (5) a' b' c' s HCI 3 IC (R ) IC IC IC 4 (kaike Information Criterion) [,7,] BIC[3] 4 : PD IC 4. IC SH IC PD Fitts ISO 94-9 IC [] [3] ( 6 0 ) OS indows CE PD 4 ˆ IC IC= l ( θ ) + ( parameters' number ) 57 mm ( ) x 79 mm () x = N(logπσˆ + ) + ( m+ ) 8 ()00 g 480 x 60 pixels ( pixel = 0.4 mm) l( θˆ ) Java ˆ N N l( θ) = logπσˆ N PD m ( ) ˆσ N σˆ = ei N ( e i ) PD i= IC IC
IC IC Sannon MT = 36.46+ 9.99log ( + ) 47465 e MT = 53.5+ 53.05log ( + ) 47859 e SH MT = e log ( + ) log ( ) 5.4034 0.7075-0.000 P 46077 Fitts ((4)) MT = 97.39+ 75.3log ( + ) (6) IC 3897 0 IC 39078 Fitts 3 x 3 IC 3 : 0, 0, 40 pixels (.4, 4.8, 9.6 mm) SH : 00, 00, 300 pixels (4, 48, 7 λ = 0 mm) 5.734 0.6365-0.0755 MT 9 = e log ( ) log ( ) + (7) P 30 IC 5 37696 90 pixels 3 4 ln( MT ) - ln( ID s ) 30 c 3( ) x (-0.0755) 3( ) x 9( ) x ( ) = 33 P IC SH IC (37696) SH 33( )-8( )=304 Fitts λ 4. IC SH λ 0 3 ((5))Sannon ((4) ) e IC IC 5 IC IC IC ( ln( MT ) IC)( [] ) [7]
IC 5 : TBLET PC 3 5. λ = IC λ SH PD 500 Fitts 000 MT MT 500 000 500 0 500 000 500 000 500 0 0 3 4 5 6 ID Sannon 0 0.5.5.5 3 3.5 4 IDe 3e 9 8 7 6 9 3 38 indows XP FMV STYLISTIC tablet PC Java cm x 5.6 cm (0.055 mm/pixel) PD 3 x 3 :, 36, 7 pixels : 0, 360, 840 pixels 760 pixels 9 Fitts [6] 3( ) x 3( ) x 3( ) x ( ) x ( ) = 3564 8 3564-8=3536 ln(mt) 5 4 3 0 0 0. 0.4 0.6 0.8..4.6.8 ln(ids) 4SH ( λ =0) 5. IC IC SH IC (46077)
λ SH IC λ =0 λ = λ = λ =3 37696 37689 3769 37694 46077 46037 4603 46039 SH e Sannon (%) (%) (%) (%) (%) (%) 47.8 5. 54.9 45. 55. 44.9 5.8 47. 58.9 4. 59.7 40.3 λ 5 SH IC 3 λ = IC (+) (+3) λ 6 PD SH IC Tablet PC IC (e) Sannon PD PD + + 3 IC 0.5 0.5 IC IC + 6.% 0.9% PD 5λ = tablet PC SH IC IC SH λ λ λ IC 3 IC λ = λ = IC λ
Sumin Zai IBM lmanden Researc Center. ccot, J., Zai, S., More tan dotting te I s 3 Foundations for crossing-based interfaces. Proc. CHI 00, CHI Letters 4(), 73-80, 00. 4. kaike, H. new look at te statistical model identification, IEEE Trans. uto. Control, C-9, 76-73, 974. Ln(MT) 3. ISO94-9: Ergonomic design for office work wit MT Ln(MT) visual display terminals (VDTs) Part 9: Requirements for non-keyboard input devices. 000, International 4 Standardization Organization. 4. Everitt, B.S. Te Cambridge Dictionary of Statistics, SH Cambridge University Press, London, 998. 5. Fitts, P.M. Te information capacity of te uman motor system in controlling te amplitude of movement. Journal of Experimental Psycology, 47. 38-39, 954. 6. Fitts, P.M. and Radford, B.K. Information capacity of discrete motor responses under different cognitive sets. Journal of Experimental Psycology, 7 (4), 475-48, 966. 7. Kitagawa, G. and Gersc,. Smootness Priors Sannon nalysis of Time Series, Springer-Verlag, New York, SH 996. 8. MacKenzie, I. S., note on te information-teoretic 4% basis for Fitts' law. Journal of Motor Beavior,, 33-330,989. 9. Mackenzie, I. S. Fitts' Law as a performance model in uman-computer interaction. Doctoral dissertation. University of Toronto: Toronto, Ontario, Canada, 99. 0. Mackenzie, I. S. Fitts law as a researc and design tool in uman-computer interaction. Human-Computer Interaction, 9-39, 99. HCI. Press, S. J. Bayesian Statistics. Principles, Models, and pplications, Jon iley & Sons, New York, 989. IC. Sakamoto, T., Isiguro, M. and Kitagawa, G. kaike HCI Information Criterion Statistics, D. Reidel, Dordrect, 986. IC 3. Scwarz, G. Estimating te dimension of a model, nnals of Statistics, 6(), 46-464, 978. 4. Sannon, C.E. Matematical Teory of Communication. Te Bell System Tecnical Journ al, 7, 397-43, 63-656, 948. 5. Vinod, H.D. and Ulla,. Recent dvances in Regression Metods, Marcel Dekker, New York, 98. 6. Zai, S. On te Validity of Trougput as a Caracteristic of Computer Input. IBM Researc Report, RJ 053(008-06) ugust, 00.