270 likes | 527 Views
Parasuraman , Sheridan,& Wickens 2000. A Model for Types and Levels of Human Interaction with Automation. Introduction. The paper proposes a model that describes the types and levels of automation, and gives a basis for applying them Levels of automation High: fully automatic
E N D
Parasuraman, Sheridan,& Wickens 2000 A Model for Types and Levels of Human Interactionwith Automation
Introduction • The paper proposes a model that describes the types and levels of automation, and gives a basis for applying them • Levels of automation • High: fully automatic • Low: fully manual • Types of automation • Information acquisition • Information analysis • Decision and action selection • Action implementation
Automation • Working definition: “… a device or system that accomplishes (partially or fully) a function that was previously, or conceivably could be, carried out (partially or fully) by a human operator.” • 4-stage model of human information processing • 1. Sensory Processing: “… positioning and orienting of sensory receptors, sensory processing, initial pre-processing of data prior to full perception, and selective attention.” • 2. Perception: “… conscious perception, and manipulation of processed and retrieved information in working memory [13]. This stage also includes cognitive operations such as rehearsal, integration and inference…” • 3. Decision Making: “… decisions are reached based on such cognitive processing.” • 4. Response Selection: “… implementation of a response or action consistent with the decision choice.” • Different levels of automation can be implemented at each of these stages.
TABLE 1 & FIGURE 1LEVELS OF AUTOMATION OF DECISIONAND ACTION SELECTION
Automation Continued • Acquisition • Low: Moving sensors & scanning • Moderate: Organizing & prioritizing information • High: Information filtering • Analysis • Low: Prediction & extrapolation • Moderate: Integration & emergent features • High: Information management & contextual summary • Decision • Low: Providing action suggestion • Moderate: Selecting an action that operator can veto • High: Selecting & performing an action • Action • Low: Performing operator selected actions • Moderate: Operator “hands-off” control to the automation • High: Automation monitors task and executes operations automatically • Adaptive automation • Dynamic changing of levels of automation for different situations
Framework for Automation Design Fig. 3. Flow chart showing application of the model of types and levels of automation. For each type of automation (acquisition, analysis, decision, and action), a level of automation between low (manual) and high (full automation) is chosen. This level is then evaluated by applying the primary evaluative criteria of human performance consequence, and adjusted if necessary, in an iterative as shown. Secondary evaluative criteria are then also iteratively applied to adjust the level of automation. The process is then repeated for all four types of automation.
Human Performance Consequences • Primary Criteria • Mental Workload • What levels enhance performance & which degrade? • Situation Awareness • What levels improve awareness & which diminish? • Complacency • Can different levels be used to compensate for system reliability & avoid complacency? • Skill Degradation • How can skill degradation be avoided? • Secondary Criteria • Automation Reliability • How do we determine reliability? • Cost of Outcomes • How do we determine risks & costs of automation failure & operator involvement? • Interdependence of decision & action • Application Example • Air Traffic Control (ATC) Systems
Conclusions • The model provides a framework on which decisions about automation implementation can be based. • The model can guide the automation design process.
ENDSLEY & KABER 1999 Level of automation effects on performance, situationawareness and workload in a dynamic control task
LOA (Level Of Automation) Taxonomy • Sheridan and Verplanck (1978) • (1) human does the whole job up to the point of turning it over to the computer to implement; • (2) computer helps by determining the options; • (3) computer helps to determine options and suggests one, which human need not follow; • (4) computer selects action and human may or may not do it; • (5) computer selects action and implements it if human approves; • (6) computer selects action, informs human in plenty of time to stop it; • (7) computer does whole job and necessarily tells human what it did; • (8) computer does whole job and tells human what it did only if human explicitly asks; • (9) computer does whole job and decides what the human should be told; and • (10) computer does the whole job if it decides it should be done, and if so, tells the human, if it decides that the human should be told. • Endsley (1987) • (1) manual control - with no assistance from the system; • (2) decision support - by the operator with input in the form of recommendations provided by the system; • (3) consensual artificial intelligence (AI) - by the system with the consent of the operator required to carry out actions; • (4) monitored AI - by the system to be automatically implemented unless vetoed by the operator; and • (5) full automation with no operator interaction. • 10 level taxonomy of LOA • (1) Manual Control (MC) • (2) Action Support (AS) • (3) Batch Processing (BP) • (4) Shared Control (SHC) • (5) Decision Support (DS) • (6) Blended Decision Making (BDM) • (7) Rigid System (RS) • (8) Automated Decision Making (ADM) • (9) Supervisory Control (SC) • (10) Full Automation (FA)
Experimental Task • Participants tracked boxes of different sizes & colors moving towards the center of the screen & attempted to keep them from colliding with the or each other. Different penalties & rewards were associated with different colors & sizes. Operators were also subjected to automation failures during the task. • LOA was manipulated for different operators & task performance during normal operation & automation failure, operator SA (Measured at three levels of engagement), & operator mental workload were recorded.
Conclusions • LOAs that combine human generation of options with computer implementation produce superior performance under normal conditions. • Participants benefited most from computer assistance in implementation, computer generation and selection was not superior to operators performance. • Computer guidance in operation selection produced the worst performance. • LOAs that had operators focused on future system states where most disruptive during automation fa