Article summary with Information Processing (Chapter 4) by Wickens & Carswell - 2012
- Introduction
- Three approaches to information processing
- The classic stage-based approach
- The ecological approach
- The cognitive engineering approach
- Selecting information
- Selective attention
- Focused attention
- Divided attention
- Visual search
- Perception and data interpretation
- Judgments of Distance and Size in Three-Dimensional Space
- Comprehension and cognition
- Metacognition
- Action selection
- Modality
- Multiple-task performance
Introduction
In many situations, humans interact with systems. During these interactions, the operator must perceive information and transform information into different forms. Sometimes these transformations lead to errors. Understanding these transformations and thus understanding information processing, is important for predicting and modeling human-system interactions.
Three approaches to information processing
There are three distinct approaches to information processing: the classic stage-based approach; the ecological approach and cognitive engineering or ergonomics.
The classic stage-based approach
In this approach, the digital computer is used as a metaphor to human behavior. Information is seen as passing through a number of discrete stages. So, there is a distinction between a perceptual stage and a stage of execution and action, which is based on the morphological distinctions between perceptual and motor cortex. Proof for this approach comes from the fact that different tasks and environmental factors have a different influence on different stages. Within this approach, processing does not always start at stage 1: sometimes processing starts when someone gives a response.
The ecological approach
This approach places more emphasis on the integrated flow of information through the human rather than making a distinction between stages. It also emphasizes the interaction between humans and the environment. This approach is most relevant to describing human behavior in interaction with the natural environment, so it is used most when designing controls and displays that mimic characteristics of the natural environment.
The cognitive engineering approach
This approach is a hybrid of the stage-based and ecological approach. On the one hand, it is based on a very careful understanding of the environment and tasks constraints within which an operator works. On the other hand, this approach places great emphasis on modeling and understanding the knowledge structures that expert operators have of the domain.
Selecting information
Broadbent's book lead to that human information processing is now seen as part of a filtering process. This filtering happens through mechanisms of human attention. Attention has three different modes: selective attention, focused attention and divided attention.
Selective attention
Selective attention refers to how attention is focused on a particular object in the environment for a certain period of time. It is influenced by four factors: salience, effort, expectancy and value. So, selective attention dictates where attention is given to.
Focused attention
Focused attention is used to maintain processing of the desired sources and avoid the distracting influence of potentially competing stimuli.
Divided attention
This is the ability to process more than on attribute or element of the environment at a given time.
Visual search
When people are looking for something in a cluttered environment (for instance when they are looking for a sign by the roadway), they use selective and focused attention as well as discrimination. Visual search models are used to predict the time that is required to find a target. These predictions can be very important for safety and productivity.
The most simple model of visual search is the 'serial self-terminating model'. In this model, it states that search space is filled with items of which most are nontargets (so, distractors). The mean time to find a target is modeled to be RT = NT/2. N is the number of items in the space and T is the time that is needed to examine each item and determine that it is not a target before moving on to the next. This model is influenced by three factors: bottom-up parallel processing, top-down processing and target familiarity. Bottom-up processing is about that for example all targets are 'highlighted', so that searching for these targets is easier. Top-down processing is about how the operator's knowledge or expectations influence the information processing. For example, location expectancy will create search strategies that scan the most likely locations first. Another influence is the expectancy of whether a target will be present or not. This is called the 'target prevalence rate'. A third factor that influences visual search is target familiarity: this means that repeated exposures to the same consistent target can speed the search for that target and reduce the likelihood that the target may be missed.
Perception and data interpretation
The Signal Detection Theory
When designing displays, it is very important that critical targets must be detectable in the environment. However, assuring this detectability can be difficult. It is often the case that changes in a scene are missed. However, sometimes it is also the case that people respond as if they saw something, while there was no target. This is called a false alarm. The SDT provides a framework for describing the processes that can lead to both types of errors.
Expectancy, context and identification
Prior knowledge can also influence the ability to identify enormous numbers of objects. For example, it seems that objects and attributes are recognized more quickly when they are embedded in consistent contexts, rather than when they are presented alone or in different, inconsistent contexts. For example, words are more easily identified when they are embedded in sentences, compared to when words are presented alone.
Judgments of Two-Dimensional Position and Extent
Spatial judgements that are required to read even everyday graphs, are prone to systematic distortions. A few examples of these distortions are that people overestimate the values that are represented in bar graphs; perceptual flattening of line graphs with respect to the y-axis, which results in larger underestimations of the represented data as the reader follows the line from its origin; cyclic patterns of bias in estimations of part- whole relationships that are dependent on the number of available reference points on the graphs; distance distortions between cursor locations and target/icon locations induced by the shape of the cursor.
Judgments of Distance and Size in Three-Dimensional Space
There are five kinds of cues that help during perception:
When making judgements in spaces, human perception depends on different cues that provide information about the relative or absolute distance from the viewer. Many of these cues are called pictorial cues, because they cues can be used to generate the impression of depth in 2D pictures.
Next to pictorial cues, there are five cues that have to do with characteristics of the viewer: Motion parallax: this refers to that objects moving at a constant speed across the frame will appear to move an greater amount if they are closer to an observer than they would if they were at a greater distance. Binocular disparity: this refers to the difference in viewpoint of the two eyes. Stereopsis: this is the use of binocular disparity to perceive depth (think about 3D displays). Accommodation and binocular convergence: these cues result from the natural adjustment of the eyes that is needed to focus on specific objects at different distances.
Comprehension and cognition
Working Memory limitations
There is a limited number of ideas, sounds and images that we can maintain and use in our mind. For example, the items in the working memory are lost when they are not repeated. This is called the memory span. Baddeley developed a four-part model of the working memory, which includes two temporary storage systems: the phonological loop and visuospatial sketchpad. These subsystems are used by a central executive, which manipulates the information in these stores and creates multimodal representations of coherent objects. These representations are then held in an episodic buffer.
Knowledge about the working memory has some implications for design:
Whenever possible, try to avoid codes that are too long for the memory capacity; when it is necessary to use codes that exceed this limits, use methods such as parsing material into lower units (chunking); because information from working memory is lost after a few seconds, systems should be designed in such a way that they can use the information (think about voice menu systems: in these cases, the users should be able to make a choice immediately); the need to scan should be minimized if a person must hold spatial information in the sketchpad; avoid the need to transfer information from one subsystem into the other before further transformations or integrations can be made; if working memory subsystems are updated too rapidly, old information may interfere with new; interference in working memory is most likely when-to-be remember information is similar in either meaning or sound; the capacity of working memory varies between people.
Planning and problem solving
In contrast with cognitive activities that are heavily driven by information the environment, the information-processing tasks of planning and problem solving are more dependent on the interplay between information that is available in the long-term memory and information-processing transformations carried out in working memory.
Planning
Planning can depend on two types of cognitive operations: planners may depends on scripts that they have stored in their long-term memory which are based on past experience; planning may involve guess work and some level of mental stimulation of future activities.
Problem solving, diagnosis and troubleshooting
These three activities are related to each other: they all have in common that there is a goal to be obtained by the human operator, that information to achieve that goal is currently missing and that some physical action or mental operation must be taken to seek these entities.
Metacognition
The term metacognition refers to a person's knowledge about his or her own cognitive processes and the use of this information to regulate performance. Education is the most active area of research on metacognition: researchers have looked at how students' beliefs about their own information-processing capabilities influence learning strategies and ultimate academic success.
Most researchers make a distinction between metacognitive knowledge and metacognitive control processes.
Action selection
In the stage of action selection and execution, it is important to look at the speed with which information is processed from perception to action. This speed is described in 'bandwidth': the amount of information processed per unit time. Units of information is described in bits.
Findings related to action selection
Response times for either rule-or skill based behavior is longer when there are more possible choices.
When people do not expect certain stimuli, they may respond more slowly.
Practice leads to that frequent events will be responded to more rapidly, but also expertise (and thus practice) in something may lead to less speedy processing for rare events, compared to novice events.
Spatial compatibility
The spatial compatibility of a display influences speed and accuracy of the control respones. One relates to the location of the control relative to the display and the other relates to how the display reflects control movement.
Modality
Machine systems are progressively more becoming executed by voice. There are three characteristics of voice control in the context of information processing: voic options allow more possible responses to be given in a shorter period of time compared to hand-control; voice options represent more compatible ways of transmitting symbolic or verbal information; voice options are valuable in environments when the eyes or hands are otherwise engaged.
Multiple-task performance
In multiple-task environments, there is a distinction between three different modes of multiple-task behavior: perfect parallel processing, degraded concurrent processing and strict serial processing. Perfect parallel processing means that two (or more) tasks are performed concurrently as well as either is performed alone, degraded concurrent processing means that both tasks are performed concurrently but one or both suffers, and strict parallel processing means that only one task is performed at a time.
When two tasks are similar, this may lead to confusion. Easier tasks are more likely to be performed perfectly compared to more difficult tasks. Also, when the two tasks are different, then this may lead to better performance.
- 1605 reads
Add new contribution