Modeling Human–System Interaction

Philosophical and Methodological Considerations, with Examples

Thomas B. Sheridan

• Publisher: Wiley
• Year: 2016
• Print ISBN: 9781119275268
• eBook ISBN: 9781119275282
• Edition: 1
• Topic: Technology & Engineering
• Subtopic: Electrical Engineering & Telecommunications
• Index: Technology & Engineering

This book presents theories and models to examine how humans interact with complex automated systems, including both empirical and theoretical methods.

• Provides examples of models appropriate to the four stages of human-system interaction
• Examines in detail the philosophical underpinnings and assumptions of modeling
• Discusses how a model fits into "doing science" and the considerations in garnering evidence and arriving at beliefs for the modeled phenomena

Modeling Human-System Interaction is a reference for professionals in industry, academia and government who are researching, designing and implementing human-technology systems in transportation, communication, manufacturing, energy, and health care sectors.

1
KNOWLEDGE

GAINING NEW KNOWLEDGE

Knowledge can be acquired by humans in many ways. Surely, there are also many ways to classify the means to acquire knowledge. Here are just a few ways.

One’s brain can acquire knowledge during the evolutionary process by successive modifications to the genes. That finally results in fertilization of egg by sperm and the gestation process in the mother. Certainly, all this depends on the sensory–motor “operating system” software that makes the sense organs and muscles work together. But evolution also plays at the level of higher cognitive function. As Noam Chomsky has shown us (Chomsky, 1957), much of the syntactic structure of grammar is evidently built in at birth. What knowledge we acquire after birth is a function of what we attend to, and what we attend to is a function of our motivation for allocating our attention, which ultimately is a function of what we know, so knowledge acquisition after birth is a causal circle.

Learning has to do with how we respond to the stimuli we observe. Perhaps, the oldest theory of learning is the process of Pavlovian (classical) conditioning, where a stimulus, originally neutral in its effect, becomes a signal that an inherently significant (reward or punishment) unconditioned stimulus is about to occur. This results only after multiple pairings, and the brain somehow remembers the association. The originally neutral stimulus becomes conditioned, meaning that the person (or animal) responds reflexively to the conditioned stimulus the same as the person would respond to the unconditioned stimulus (e.g., the dog salivates with the light or bell).

A different kind of learning is Skinnerian or operant conditioning (Skinner, 1938). This is where a voluntary random action (called a free operant) is rewarded (reinforced), that association is remembered, and after sufficient repetitions, the voluntary actions occur more often (if previously rewarded). Operant learning can be maintained even when rewards are infrequently paired with the conditioned action.

There are many classifications of learning (http://www.washington.edu/doit/types‐learning). Bloom et al. (1956) developed a classification scheme for types of learning which includes three overlapping domains: cognitive, psychomotor, and affective. Skills in the cognitive domain include knowledge (remembering information), comprehension (explaining the meaning of information), application (using abstractions in concrete situations), analysis (breaking down a whole into component parts), and synthesis (putting parts together to form a new and integrated whole).

Gardner (2011) developed a theory of multiple intelligences based upon research in the biological sciences, logistical analysis, and psychology. He breaks down knowledge into seven types: logical–mathematical intelligence (the ability to detect patterns, think logically, reason and analyze, and compute mathematical equations), linguistic intelligence (the mastery of oral and written language in self‐expression and memory), spatial intelligence (the ability to recognize and manipulate patterns in spatial relationships), musical intelligence (the ability to recognize and compose music), kinesthetic intelligence (the ability to use the body or parts of the body to create products or solve problems), interpersonal intelligence (the ability to recognize another’s intentions and feelings), and intrapersonal intelligence (the ability to understand oneself and use the information to self‐manage).

Knowledge can be public, where two or more people agree on some perception or interpretation and others can access the same information. Or it can be private, where it has not or cannot be shared. The issue is tricky, and that is why modelability is proposed as a criterion for what can be called public knowledge. Two people can look at what we call a red rose, and agree that it is red, because they have learned to respond with the word red upon observing that stimulus. But ultimately exactly what they experienced cannot be shared, hence is not public.

We can posit that some learning is simply accepting, unquestioningly, information from some source because that source is trusted or because the learner is compelled in some way to learn. We finally contrast the aforementioned models to learning by means of the scientific method, which is detailed in the following text. Critical observation and hypothesizing are followed by collection of evidence, analysis, logical conclusions, and modeling to serve one’s own use or to communicate to others.

SCIENTIFIC METHOD: WHAT IS IT?

How to determine the truth? Science has its own formal model for this. The scientific method is usually stated as consisting of nine steps as follows:

1. Gather information and resources (informal observation).
2. Question the relationships between aspects of some objects or events, based on observation and contemplation. An incipient mental model may already form in the observer’s head.
3. Hypothesize a conjecture resulting from the act of questioning. This can be either a predictive or an explanatory hypothesis. In either case, it should be stated explicitly in terms of independent and dependent variables (causes and effects).
4. Predict the logical consequences of the hypothesis. (A model will begin to take shape.)
5. Test the hypothesis by doing formal data collection and experiments to determine whether the world behaves according to the prediction. This includes taking pains to design the data‐taking and the experiment to minimize risks of experimental error. It is critical that the tests be recorded in enough detail so as to be observable and repeatable by others. The experimental design will have a large effect on what model might emerge.
6. Analyze the results of the experiment and draw tentative conclusions. This often involves a secondary hypothesis step, namely exercising the statistical null hypothesis. The null hypothesis is that some conjecture about a population of related objects or events is false, namely that observed differences have occurred by chance, for example, that some disease is not affected by some drug. Normally, the effort is to show a degree of statistical confidence in the failure and thus rejection of the null hypothesis. In other words, if there is enough confidence that the differences did not occur by chance, then the conjectured relationship exists.
7. Draw formal conclusions and model as appropriate.
8. Communicate the results, conclusions, and model to colleagues in publication or verbal presentation, rendering the model in a form that best summarizes and communicates the determined relationships.
9. Retest and refine the model (frequently done based on review and critique by other scientists).

FURTHER OBSERVATIONS ON THE SCIENTIFIC METHOD

The scientific method described earlier is also called the hypothetico‐deductive method. As stated, it is an idealization of the way science really works, as the given scientific steps are seldom cleanly separated and the process is typically messy. Often experimentation is done in order to make observations that provoke additional observations, questions, hypotheses, predictions, and rejections or refinements of the starting hypothesis. Especially at the early observation stage, the process can be very informal. One of the writer’s students used to say that what we did in the lab was “piddling with a purpose.” Einstein is said to have remarked that the most important tool of the scientist is the wastebasket.

Philosopher statesman Francis Bacon (1620) asserted that observations must be collected “without prejudice.” But as scientists are real people, there is no way they can operate free of some prejudice. They start with some bias as to their initial knowledge and interests, their social status and physical location, and their available tools of observation. They are initially prejudiced as to what is of interest, what observations are made, and what questions are asked.

Philosopher Karl Popper (1997) believed that all science begins with a prejudiced hypothesis. He further asserted that actually a theory can never be proven correct by observation, but it can only be proven incorrect by disagreement with observation. Scientific method is about falsifiability. That is the basis of the null hypothesis test in statistics. (But, of course, the falsifiability is itself subject to statistical error; one can only reject the null hypothesis with some small chance of being wrong.) The American Association for the Advancement of Science asserted in a legal brief to the U.S. Supreme Court (1993) that “Science is not an encyclopedic body of knowledge about the universe. Instead, it represents a process for proposing and refining theoretical explanations about the world that are subject to further testing and refinement.”

Historian Thomas Kuhn (1962) offered a different perspective on how science works, namely, in terms of paradigm shifts. Whether in psychology or cosmology, researchers seem to make small and gradual refinements of accepted models, until new evidence and an accompanying model provokes a radical shift in paradigm, to which scientists then adhere for a time. When a new paradigm is in process of emerging the competition between models and their proponents can be fierce, even personal (who discovered X first, who published first, whose model offers the best explanation). We also must admit that search for truth is not the only thing that motivates us as scientists and modelers. We are driven by ambition for recognition from our peers as well as by money.

The idea of reproducible observability deserves emphasis. Having to deal with observables is the most critical factor in an epistemological sense (what we know). This is because it distinguishes what may be called truth based on scientific evidence that is openly observable from experiences that are not observable by others (e.g., personal testimony and anecdotal evidence). Observability also comes into play for what are called mental models.

Mental models can be called models of a sort, but being private they are not subject to direct observation by other people. Experiments in psychophysics, where subjects make verbal category judgments or button‐push responses to physical stimuli of sound, light, and so on, are regarded as conforming to scientific method. This is because the human is making a direct mechanical response to a given stimulus, such as pushing a button, not having to articulate in arbitrary words what the human is thinking. However, when subjects are asked to explicate in their own words their mental models of how they believe something works, or what are cognitive steps of a particular task as might be asked of subject matter experts, there is no external physical reference; scientific method here is more challenging. Of course, there must be repeatability or aggregation of the results from many subjects. Observability clearly is a challenge for modeling.

While social scientists often point out that humans have a predilection for reaffirming the status quo, science nevertheless is a truth system committed to change, as warranted, rather than preservation. But while science actively pursues possibilities of change, the null hypothesis testing by its very nature demands a significant level of statistical confidence in order to reject the null hypothesis that there is no real change (that there is only random difference between the hypothesized variant and the control).

The scientific method is a method wherein inquiry regards itself as fallible and purposely probes, criticizes, corrects, and improves itself. This universally accepted attribute stands in sharp contrast to religious and political traditions around the world. Science is the one human endeavor that has p...

Table of contents

• COVER
• TITLE PAGE
• TABLE OF CONTENTS
• PREFACE
• INTRODUCTION
• 1 KNOWLEDGE
• 2 WHAT IS A MODEL?
• 3 IMPORTANT DISTINCTIONS IN MODELING
• 4 FORMS OF REPRESENTATION
• 5 ACQUIRING INFORMATION
• 6 ANALYZING THE INFORMATION
• 7 DECIDING ON ACTION
• 8 IMPLEMENTING AND EVALUATING THE ACTION
• 9 HUMAN–AUTOMATION INTERACTION
• 10 MENTAL MODELS
• 11 CAN COGNITIVE ENGINEERING MODELING CONTRIBUTE TO MODELING LARGE‐SCALE SOCIO‐TECHNICAL SYSTEMS?
• APPENDIX
• REFERENCES
• INDEX
• END USER LICENSE AGREEMENT