Download pptx - Ecological Interface Design (EID)

UI/CM LAB

Ecological Interface Design (EID)

I. INTRODUCTION an attempt to extend the benefits of direct manipulation interfaces

(DMI) to complex work domains

A. Unanticipated Events thee broad areas for events1) familiar events – skills to deal with these events2) unfamiliar, but anticipated events – infrequent3) unfamiliar and unanticipated events unanticipated events are the major cause of life-threatening accidents

mistakes (errors in intention) rather than slips (errors in execution)for routine events

UI/CM LAB


II. PROBLEM FORMULATIONA. Fundamentals an interface a control system control theory1. the Law of Requisite Variety states that complex systems require com-

plex controllers2. physical systems can be described by a set of constraints3. every good controller must be, or possess, a model of the system it is

controlling

UI/CM LAB


B. The Structure of the Design Problem two questions pertinent to interface design

• Abstraction hierarchy as a psychologically relevant form of representing the constraints in a work domain in a way that operators to cope with unanticipated events

• SRK taxonomy as a useful framework for describing the various mechanisms that people have for processing info

• domain and operator as the organism-environment reciprocity

UI/CM LAB


III. THE ABSTRACTION HIERARCHYA. What Kind of Hierarchy? belongs to the class of stratified hierarchy by Mesarovic et al. (1970) a means-end relationship between levels (explicitly goal-oriented na-

ture) the AH not a specific representation but rather a framework for develop-

ing representations for various work domains five levels of constraints for process control systems – functional pur-

pose, abstract function, generalized function, physical function, physical form

an informational basis for coping with unanticipated events and a psy-chologically valid representation for problem-solving

UI/CM LAB


UI/CM LAB


B. Coping with Unanticipated Events: A Historical Overview in the 1960’s at Risø National Lab in Denmark human operator plays a key role in overall system reliability and safety accident-causing errors by human operators with unfamiliar situations

not anticipated by designers Engineering analysis of the control requirements posed by unantici-

pated events the AH as a framework for identifying and integrating the set of goal rel-

evant constraints that are operating in a given work domain higher levels represent relational info about system purpose, whereas

the lower levels represent more elemental data about physical imple-mentation

UI/CM LAB


D. Psychological Relevance engineering perspective for coping with unanticipated events higher levels are less detailed than lower levels makes complex sys-

tems look simpler explicitly goal-oriented means-end relation an efficient form of

search1. possible to meaningfully map problem solving protocols onto an AH rep-

resentation of the domain2. subjects’ problem solving trajectories would begin at a high level of ab-

straction and gradually focus in on lower levels, thereby exploiting the goal-related constraint provided by the hierarchy

UI/CM LAB


IV. MULTIPLE LEVELS OF COGNITIVE CONTROL three levels of info (signals, signs, or symbols) vs. three levels of cogni-

tive control SBB (automated behavior patterns), RBB (a set of cue-ac-tion mappings), and KBB (PS operations on a symbolic representation)

A. The Power of Perception three levels of cognitive control into two general categories1. perceptual processing (SBB, RBB) – fast, effortless, parallel2. analytic PS based on a symbolic representation (KBB) – slow, labori-

ous, serial, more error prone due to WM

UI/CM LAB


two characteristics of complex work domain1. highly skilled and extensive experience in controlling the system2. interface design needs specific application – generality not important make perceptual processing an attractive possibility empirical evidence? Brunswick (1956), Hammond et al. (1987) perception can be very effective not always leads to superior perfor-

mance but the conditions characteristics of complex work domains are propitious for perceptual processing

UI/CM LAB


B. The Propensity for Perceptual ProcessingTwo Examples Klein (1989) – a series of naturalistic DM in the domain of firefighting,

military operations, engineering design; nonroutine events Surprisingly, experts often relies on recognitional DM Mental effort – recognition mode less taxing than the analytic mode Effectiveness – with their experience quickly generate plausible ac-

tion alternatives than the complete set of possible alternatives Appropriateness -- recognitional DM much quicker than analytical

Kirlick (1989) – complex, supervisory control task Expert behavior almost exclusively on perception and action

UI/CM LAB


More Examples and What Can Go Wrong Hollnagel (1981) – surface/ deep control in process control Fischoff et al. (1978), Smith (1989) in management DM

C. Skill and Task Effects Level of skill of the operator and the level of complexity of the task de-

mands Necessary to understand how the levels are related and what the activi-

ties associated with each other

UI/CM LAB


E. Implications for Interface Design

SBB in the form of time-space signals, RBB by familiar perceptual forms (signs), KBB by meaningful structures (Symbols)

UI/CM LAB


V. ECOLOGICAL INTERFACE DESIGNA. The Principles1) SBB – To support interaction via time-space signals, the operator

should be able to act to directly on the display, and the structure of the displayed information should be isomorphic to the part-whole structure of movements

2) RBB – Provide a consistent one-to-one mapping between the work do-main constraints and the cues or signs provided by the interface no consistent mapping between the perceptual cues and the con-

straints that govern the process behavior procedural traps: novel situations where operators rely on their normal rule set, but without the usual success

UI/CM LAB


Unique and consistent mapping between constraints and cues1. mental economy – RBB less than KBB2. Can exhibit what looks like KBB by relying on RBB

3) KBB – Represent the work domain in the form of an abstraction hierar-chy to serve as an externalized mental model that will support knowl-edge-based problem solving

B. Limitations Three issues pertaining to the use of AH1. designers’ knowledge of the constraints governing the system2. Robustness – empirical research needed3. Limitations due to sensor technology4. generalizability

UI/CM LAB


VI. WHAT IS THE CONTRIBUTION OF EID?A. Communicating the Information to the OperatorDirect Manipulation Interfaces: DMI allow users to directly act on what they see in the display but lack

of the explanation of human information processing capability – SRK framework

Object Displays: Mapping the higher order perceptual relationships onto goal-relevant

variables Directly relevant to EID principle 2 – to support RBB, the perceptual

cues (signs) should directly specify the process constraints

UI/CM LAB


Technology-Driven Display Design: Computer graphics for building interfaces for complex technical systems

– mimic or schematic diagrams EID is top-down while technology-based is bottom-up

B. Representing the Complexity in the DomainOperator Function Model: Another formalism as an alternative to the AH OFM (Mitchell and Miller, 1986) – a discrete control model1) What data should be displayed?2) How should those data be organized into screens?3) How should context sensitivity be built into the display?4) How can information be presented at various levels of detail?

UI/CM LAB


Mappings are nondeterministic Miller (1982) – structural vs. behavioral representation

A structural representation is one in which the structures that de-fine the system are defied directly in some set of objects (AH) – complex systems where unanticipated events are the biggest threat to system safety

a behavioral representation is a representation of a dynamic sys-tem whose elements consist of system behavior (OFM) – situa-tions where operators are required to dynamically select the rele-vant subset of data to carry out predictable tasks

UI/CM LAB


VII. EMPIRICAL EVALUATION DURESS (Dual Reservoir System Simulation) a thermal-hydraulic

process simulation Physical/functional (P+F) interface and physical (P) interface Functional variables to the higher levels of the AH the P+F interface superior diagnosis performance to the P inter-

face, primarily for experts The higher-order functional info in the P+F interface is important

for diagnosis, justifying for including higher levels of the AH in an interface

Superiority of the P+F interface in the memory task