ch_4 selection and control of action
DESCRIPTION
Handbook of Human Factors and Ergonomics, Third Edition. CH_4 Selection and Control of Action. Robert W. Proctor Kim-Phuong L. Vu Edited by Gavriel Salvendy Lim, Soo Yong. Contents. Introduction Selection of Action Methods Action Selection in Single-Task Performance - PowerPoint PPT PresentationTRANSCRIPT
Korea Univ.Division Information Management Engineering
UI Lab.
2010 UI LAB Seminar
CH_4Selection and Control of Action
Robert W. ProctorKim-Phuong L. Vu
Edited by Gavriel Salvendy
Lim, Soo Yong
Handbook of Human Factors and Ergonomics, Third Edition
Korea Univ.Division Information Management Engineering
UI Lab.
2010 UI LAB Seminar
Contents • Introduction• Selection of Action
– Methods– Action Selection in Single-Task Performance– Action Selection in Multiple-Task Performance
• Motor Control– Methods– Control of Selection– Coordination of Effectors– Sequencing and Timing of Action– Motor Learning and Acquisition of Skill
• Summary and Conclusions
Korea Univ.Division Information Management Engineering
UI Lab.
2010 UI LAB Seminar1. Introduction
• Research on selection and control of action has developed con-temporaneously with that on human factors.– Hick-Hyman law and Fitts’ law are the few well-established quantitative laws of
behavior
• Paul M. Fitts– He made many empirical and theoretical contributions to knowledge concern-
ing selection and control of action.• Fitts’s law• the principles of stimulus-response(S-R) compatibility
• The relation between perception and action is currently a very ac-tive area of research in psychology and associated fields.
3/16
Korea Univ.Division Information Management Engineering
UI Lab.
2010 UI LAB Seminar2. Selection of Action
2.1 Methods• Selection of action is most often studied in choice-reaction tasks.
– Response time (RT) is recorded as the primary measure and error rate as a sec-ondary measure.
• Primary variables affecting the duration of action process– S-R uncertainty, S-R compatibility, response precuing, sequential dependencies
• Information about the nature of action selection– mean RT, RT distribution, percentage of error, specific types of errors, indicators
of brain functions• Speed-accuracy trade-off (figure 1)
– one well-established principle of performance in choice-reaction tasks• Choice RT methods
– not only can be used to examine action selection under single-task performance,– but also under conditions in which two or more task sets must be maintained,
• actions to each must be performed concurrently• or the person is required to switch between the various tasks periodically.
4/16
Korea Univ.Division Information Management Engineering
UI Lab.
2010 UI LAB Seminar2. Selection of Action
2.2 Action Selection in Single-Task Performance2.2.1 Uncertainty and Number of Alternatives: Hick-Hyman Law
• Hick-Hyman Law (figure 2)– interested in whether effects of S-R uncertainty could be explained in terms of in-
formation theory– showing a systematic increase in choice RT as the number of S-R alternatives in-
creased.
– In both Hick and Hyman’s study, RT increased as a logarithmic function of the average amount of information conveyed by a stimulus.
– the cost associated with high event uncertainty can be reduced by• using highly compatible display-control arrangements• giving the operators training on the task
5/16
Korea Univ.Division Information Management Engineering
UI Lab.
2010 UI LAB Seminar2. Selection of Action
2.2 Action Selection in Single-Task Performance (Cont.)2.2.2 Stimulus-Response Compatibility
• Spatial Compatibility (figure 3)– The responses were fastest and most accurate when the display and control
configurations corresponded spatially than when they did not. (Fitts and Seeger, 1953)
– Simon effect (figure 4)• spatial correspondence not only benefits performance when stimulus loca-
tion is relevant to the task but also when it is irrelevant• Accounts of S-R Compatibility
– S-R compatibility effects attribute them to two factors• direct, or automatic, activation of the corresponding response• intentional translation of the stimulus into the desired response
• Dimensional Overlap– Kornblum introduced the term dimensional overlap to describe stimulus and re-
sponse sets are perceptually or conceptually similar.• Action Goals and Structural Correspondence (figure 5)
– S-R Compatibility effects are determined largely by action goals and not the ac-tual physical responses.
6/16
Korea Univ.Division Information Management Engineering
UI Lab.
2010 UI LAB Seminar2. Selection of Action
2.2 Action Selection in Single-Task Performance (Cont.)2.2.3 Sequential Effects
• Repetition benefit– increases in size as the number of S-R alternatives becomes larger and is greater
for incompatible S-R mappings than for compatible ones– residual activation from the preceding trial when the current trial is identical to it
and intentional preparation for what is expected on the next trial
• Negative priming (figure 6)– in the Stroop color-naming task, RT is typically longer if the relevant stimulus
value on a trial is the same as that of the irrelevant information on the previous trial
• Simon task– non-corresponding information from the previous trial can alter how the present
trial is processed– suppression/release hypothesis (Stürmer, 1999)– event-file hypothesis (Hommel, 1998; 2004)
7/16
Korea Univ.Division Information Management Engineering
UI Lab.
2010 UI LAB Seminar2. Selection of Action
2.2 Action Selection in Single-Task Performance (Cont.)2.2.4 Preparation and Advance Information
• When a stimulus to which a response is required occurs unexpect-edly, the response to it will be slower than when it is expected.– RT decreases and error rate increases as the warning interval increases
• People can also use informative cues to prepare for subsets of stimuli and responses.
2.2.5 Acquisition and Transfer of Action-Selection Skill• Practice results in performance becoming increasingly automa-
tized.– Newell and Rosenbloom (1981)
– Heathcote (2000)
• Practice in spatial choice tasks involve primarily the mappings of the stimuli to spatial response codes and not to the specific motor effectors.
A : asymptotic RTB : performance time on the first trialN : the number of practice trialsβ : learning rateα : rate parameter
8/16
Korea Univ.Division Information Management Engineering
UI Lab.
2010 UI LAB Seminar2. Selection of Action
2.3 Action Selection in Multiple-Task Performance2.3.1 Task Switching and Mixing Costs1. Mixing cost2. Switch cost3. Preparation benefit4. Residual cost
• Cost associated with mixing an easy task with a more difficult one are often larger for the easier task.
• In a two-choice spatial compatibility task, the advantage for the compatible spatial mapping is eliminated when compatible and in-compatible mappings are mixed.
9/16
Korea Univ.Division Information Management Engineering
UI Lab.
2010 UI LAB Seminar2. Selection of Action
2.3 Action Selection in Multiple-Task Performance (Cont.)2.3.2 Psychological Refractory Period (PRP)
• Refers to slowing of RT for the second of two tasks that are per-formed in rapid succession
• Central Bottleneck Model (figure 7)– the most widely accepted account of the PRP effect
• EPIC– architecture consists of perceptual, cognitive, and motor components that does
not include a limit on central processing capacity– limits in the systems are attributed to the sensory and motor effectors, not to the
central processes• central limitations arise from individuals’ strategies for satisfying task de-
mands
10/16
Korea Univ.Division Information Management Engineering
UI Lab.
2010 UI LAB Seminar3. Motor Control
3.1 Methods• Some issues include the nature of movement representation, the
role of sensory feedback in movement execution, and the way in which motor actions are sequenced.
3.2 Control of Action• Open-loop vs. Closed-loop
– a movement under open-loop control can be executed quickly, without a delay to process feedback, but at a cost of limited accuracy
– closed-loop control is slower but more accurate– both types of control are often combined
3.2.1 Fitts’s Law (figure 8)
– Meyer provided the most complete account of the relation, a stochastic opti-mized-submovement model.
11/16
Korea Univ.Division Information Management Engineering
UI Lab.
2010 UI LAB Seminar3. Motor Control
3.2 Control of Action (Cont.)3.2.2 Motor Preparation and Advance Specification of Movement
Properties• Lateralized readiness potential
– movement of a limb is preceded by preparatory processes at the motor system.
• Advance specification of movement parameters– one or more parameters are precued prior to presentation of the stimulus to
which the person is to respond, the idea being that RT will decrease if those pa-rameters can be specified in advance.
– drawback : the particular patterns of results may be determined more by S-R compatibility rather than by the motoric preparation process itself
12/16
Korea Univ.Division Information Management Engineering
UI Lab.
2010 UI LAB Seminar3. Motor Control
3.2 Control of Action (Cont.)3.2.3 Visual Feedback
• Woodworth (1899)– visual feedback had no effect on performance at rates of 180 per minute or
greater– the minimum time required to process visual feedback was 450 ms
• Keele and Posner (1968)– the minimum duration for processing visual feedback is between 190 and 260
ms• Zelaznik (1983)
– feedback can be used for movements with durations of only slightly longer than 100 ms
• Proteau and Cournoyer (1992)– visual feedback decreases as a movement task is practiced– but evidence indicates that vision remains important
13/16
Korea Univ.Division Information Management Engineering
UI Lab.
2010 UI LAB Seminar3. Motor Control
3.3 Coordination of Effectors• Bimanual movements – tendency toward mirror symmetry (
figure 9)– easy to perform symmetric movements of the arms– attributed to co-activation of homologous muscles
• Bias is toward spatial symmetry and not motor symmetry– motion commands plays virtually no role in defining preferred coordination pat-
terns, in particular the symmetry tendency3.4 Sequencing and Timing of Action• The speed with a sequence of actions
– motor plans are structured hierarchically (figure 10)– the pattern of latencies was predicted by a model that assumed the memory rep-
resentation for the sequence was coded in a hierarchical decision tree• The timing of actions (Wing and Kristofferson)
– variance of the interval between responses should increase as the delay be-tween the responses increases
– the adjacent interresponse intervals should be negatively correlated
14/16
Korea Univ.Division Information Management Engineering
UI Lab.
2010 UI LAB Seminar3. Motor Control
3.5 Motor Learning and Acquisition of Skill3.5.1 Practice Schedules
• Distributed practice vs. Massed practice– massed practice depresses learning in motor tasks that require continuous
movements– for discrete tasks, massed practice may be beneficial to learning
• Contextual Interference Effect– retention and transfer of motor tasks were better when the tasks are practiced in
random order than in distinct blocks• Three types of part-task, or part-whole, practice distinguished (
figure 11)– segmentation– fractionation– simplificaiton
3.5.2 Provision of Feedback• Knowledge of results(KR), Knowledge of performance
15/16
Korea Univ.Division Information Management Engineering
UI Lab.
2010 UI LAB Seminar4. Summary and Conclusions
• For performance of the human component to be optimal,– it is necessary not only to consider how the machine should display information
regarding its states and activities to the human,– but also to take into account the processes by which the human selects and exe-
cutes actions in the sequence of the interaction
• Models of various types have been developed for various domains that provide predictions of how performance will be affected by numerous variables.
• The laws, principles, and model characteristics can be incorpo-rated into cognitive architectures such as EPIC and ACT-R/PM– that enable quantitative predictions for complex tasks of the type encountered in
much of human factors.
16/16
Korea Univ.Division Information Management Engineering
UI Lab.
2010 UI LAB SeminarQ & A
?Q & A
Korea Univ.Division Information Management Engineering
UI Lab.
2010 UI LAB SeminarFigure 1. Speed-accuracy Trade-off
4
Korea Univ.Division Information Management Engineering
UI Lab.
2010 UI LAB SeminarFigure 2. Hick-Hyman Law
5
The slope is typically shallower for highly compatible S-R pair-ings than for less compatible ones.
Korea Univ.Division Information Management Engineering
UI Lab.
2010 UI LAB SeminarFigure 3.
6
Korea Univ.Division Information Management Engineering
UI Lab.
2010 UI LAB SeminarFigure 3-1
6
Korea Univ.Division Information Management Engineering
UI Lab.
2010 UI LAB SeminarFigure 3-2
6
Korea Univ.Division Information Management Engineering
UI Lab.
2010 UI LAB SeminarFigure 4.
6
Korea Univ.Division Information Management Engineering
UI Lab.
2010 UI LAB Seminar
24
Figure 5.
음성 자극 (action goal) 이 주어지는 곳으로 움직이도록 하는데 , light 가 그 위치에 있을 수록 더 performance 가 좋다 .
6
Korea Univ.Division Information Management Engineering
UI Lab.
2010 UI LAB Seminar
25
Figure 6
Nintendo DS 의두뇌 training 게임
7
Korea Univ.Division Information Management Engineering
UI Lab.
2010 UI LAB SeminarFigure 7. Central Bottleneck Model
10
Korea Univ.Division Information Management Engineering
UI Lab.
2010 UI LAB SeminarFigure 8. Fitts’ Law
11
Korea Univ.Division Information Management Engineering
UI Lab.
2010 UI LAB SeminarFigure 9.
14
Korea Univ.Division Information Management Engineering
UI Lab.
2010 UI LAB SeminarFigure 10.
14
Korea Univ.Division Information Management Engineering
UI Lab.
2010 UI LAB SeminarFigure 11.
15