Adding Motivation to the Kolb Learning Model

By: Gary J. Salton, Ph.D., Chief R&D and CEO
Professional Communications, Inc.

INTRODUCTION
All learning involves change. Any change requires energy. Motivation describes the amount of energy that a person is willing to spend. Without motivation, learning will not occur. With high enough motivation learning cannot be stopped.

Research was conducted on five classes in three cities in Texas, Arizona and North Carolina. It involved 184 learners participating in 5 classes. The result is the identification of two markedly different kinds of motivation that required different strategies to initiate and that differ markedly in power. These were measured and integrated into the Kolb Learning Model using "I Opt" technology (see www.iopt.com for an outline of the validated technology and www.oeinstitute.org for its theoretical and operational foundations).

In addition, a new concept of "motivated learning" and a means to measure it was revealed. Applying this new concept disclosed a previously unrecognized learning issue. The concern involves a trade off in the depth and distribution of knowledge that can be controlled using the motivated learning concept and measurment.


MEASURING MOTIVATION
Human motivation has two interrelated aspects, the physical and the mental. The mental domain is controlled by rational processes These are methods that use logic, reason and/or analytical thought (Cambridge Dictionary, 2002). The “pleasure principle” of utilitarianism can be used to estimate rational motivation. People seek pleasure and avoid pain. Motivation is calculated by subtracting costs (i.e., pain) from benefits (i.e., pleasure). The more benefits exceed costs, the greater the rational motivation.

The physical element is controlled by biochemistry. “Feelings” are used to describe the emotional state that is being experienced. For example, a person can feel depressed, anxious or elated. These states can influence the attention a person is prepared to devote to a subject. They also affect a willingness to dedicate energy to learning.

Rational Motivation
Rational motivation is the “what’s in it for me” element. It was measured by asking class members two questions several days prior to the class:

Table 1
MEASURING RATIONAL MOTIVATION

Both questions use ordinal (rank ordered) scales. Subtracting ordinal scales has no crisp mathematical meaning. However, fuzzy logic inference can help. Subtracting “a little” from “a lot” does suggest that there will be something left of whatever there was a lot of. Subtracting costs from benefits seems to make logical, if not mathematical, sense.

Effort was made to insure a compatibility of scales. The two questions were asked at the same time and in the same context. The learner probably used the same mental calculus in answering both questions. This gives some justification for netting the two responses. The calculation is rough but was accepted as a measure of the rational motivation of the learner.


Emotional Motivation
Emotional motivation is a biochemical event (a feeling) not a rational calculation. It is a judgment of desirability of this “feeling.” It appears as a full-blown state that may or may not have a rational basis. It was measured by asking:

Table 2
MEASURING EMOTIONAL MOTIVATION

The scaling is again ordinal. There is no assurance that the distance between the categories is consistent. The statistics used to assess the results must be able to handle this inexactness.

Rational and emotional motivation are the two aspects that an instructor has to work with in increasing the energy the learner is willing to devote to learning. However, before proceeding it is worth testing whether they are really two different things in practice as well as theory.

Rational and Emotional Motivation are different things

The two forms of motivation can influence each other. Feeling gloomy (an emotion) can increase sensitivity to costs and diminish the value of benefits. Similarly, rational benefits can rise to a level that overwhelms whatever "feelings" you might have.

A professional mathematician, Robert C. Soltysik, was engaged to do the mathematics. He used a Multiresponse Permutation Procedure (MRPP) to test the hypothesis that the two kinds of motivation are different. The result was an estimated probability that the two types of motivation were drawn from the same distribution as p<.0000000001. Rob crosschecked the result with a Monte Carlo approximation. After 1000 iterations, it came up with p < .001 . Rob was able to state that " there is strong evidence that emotional and rational do not come from the same distribution." It is a safe bet that emotional and rational motivations are two different things in practice as well as theory.

MEASURING LEARNING

The 5 classes in the study all used the same well-established program intended to show the value of teamwork. The test was whether the class had long-term, visible effect on performance. The procedure was to wait 4 months and then ask supervisors to rate their subordinates on the improvement they actually witnessed. The following scale was used to gauge performance:

Table 3
MEASURING LEARNING

Unfortunately, the firm underwent reorganization during the study period. Some people were laid off and others no longer reported to the same supervisor. This resulted in reduction of sample size from 184 to 89 people. The loss was disappointing but enough people remained to warrant proceeding with the study.

The 89 people reported to 58 different supervisors with positions of Vice President to Assistant Account Manager. The large number and the different levels of supervisors minimize the probability of evaluator bias. The measure is probably a fair estimate of the value the class contributed to performance.


CALCULATING MOTIVATION
Comparing the motivation with the supervisor’s judgment of improvement requires that the rational and emotional scales be set to a common standard. Rob did this by subtracting the median value from each motivation data point. He then divided that total by the sum of the absolute values to create a standardized measure used in his calculations. This is a variant of a well recognized and accepted statistical technique called P-Standardization.

The broader research of which blog is a part, used multivariate optimal discriminant analysis (ODA) to predict the rank order position of each student in a class. Rob applied ODA to construct a model that used 5 variables including emotional and rational motivation. This multivariable model was able to predict the highest and lowest achievers in a class with 82.1% accuracy.

Unfortunately, Rob’s 100-iteration Monte Carlo simulation yielded a significance level of p< .15—about 1 chance in 7 that this result was produced by random events. This fell short of the academic p< .05 standard (one chance in 20) but does perform better than pure chance. The multivariable model was accepted for use in this study as a crosscheck.

Rob also created a model that compared the stand-alone effect of motivation on performance. This was called the univariable (i.e., one variable) model. The univariable model looked at how motivation affected performance while ignoring any influence other factors might have on the motivation-performance relationship.


THE RESULTS
Applying the standardized variables to the single and multivariate model produces coefficients. These measure the degree that motivation affects the performance outcome (e.g., Performance = Coefficient x Standardized Motivation Measure). Since the variables were standardized, the ratios of these coefficients serve as a measure of the relative power of both forms of motivation (see Table 4).

Table 4
RELATIVE STRENGTH OF
EMOTIONAL AND RATIONAL MOTIVATION

Both of Rob’s models give the same direction. Depending on which you accept, emotional motivation is between 3 and 6.8 times more powerful than its rational counterpart. Just to confirm these results, Rob also calculated a statistic called Kendall's tau using the univariable model. Tau is a measure of "goodness" of fit. The ratio of taus also measures relative power. The result of this was:

Table 5
RELATIVE STRENGTH USING KENDALL’S TAU

The tau calculation indicates that emotional is about 4 times as powerful as rational motivation. It is in the right direction and within the same range as the other two calculations. It is probably safe to say that emotional motivation is significantly more powerful than its rational counterpart.

This research used job improvement as a measure of learning. This means that course content will always be the single most important variable. Courses on conflict resolution will not yield any performance improvement if there is no conflict to resolve. The more relevant the course is to actual work issues, the greater will be measurable learning effect. "Needs Assessment" will always be the Holy Grail of Training & Development. There is no substitute for teaching the right course.

Motivation does not affect the way a course is taught. That is the realm of the Kolb Learning model. But motivation does affect how well Kolb will work—the higher the motivation the more that will be learned. You can read the strength of the connection between motivation and performance directly from the tau. Table 5 shows that emotional motivation accounts for about 13% (tau coefficient of .1328) of the performance improvement. Rational motivation was about 3% (tau coefficient of .0334). Next to course content (i.e., the subject being taught), emotional motivation is the single largest factor in teaching success .

It is worth noting that the 13% effect of emotional motivation is NOT based on information acquired or retained. It is a 13% improvement in the information applied to the benefit of the organization that is funding the learning. Effectively, it is a 13% improvement in productivity in the subject matter area of the class. This is an advantage of using job performance rather than conventional "tests" as a measure of teaching success.


MANAGING MOTIVATION

Motivation is not a constant. It can be changed. But to do so effectively requires that instructors, designers and management understand the differences in its two components.

Emotional

It Emotional motivation is based on biochemistry. To affect it you must effect the biochemistry of the learners. This means it has to have a physical impact. Three ways stand out as possible avenues:

Table 6
STRATEGIES FOR EMOTIONAL MOTIVATION

• Humor: Laughter is a physical response. The body moves. Chemistry is altered as this occurs. Laughter establishes rapport (a type of bond) and this increases receptivity. It is no accident that most speeches begin with a joke.

• Empathy: Observing another person’s emotional state causes a system of mirror neurons to fire in the brain. Wincing when seeing another person's mishap is evidence of this human capacity. You don't choose to feel the other person's pain, you just do. Displaying enthusiasm, passion or similar emotions tap into this aspect of human physiology. They can help setup a positive learning state. Anything that shows the emotional state occurring can trigger the mirror neurons. Instructors, students or even videos can be used to cause the mirror neurons to fire. All that is needed is a situation.

• Activity: These are planned events that involve energetic actions or movements. They can involve things like role-playing, ropes events or any other undertaking that generates an emotional response in the people participating. Unlike empathy, the direction of the emotion is harder to control when using activities. For example, a "ropes" exercise can generate fear as easily as excitement. Caution is always indicated when designing an activity intended to generate an emotional response.

Empathy, humor and activity are universal options. Initiatives focused on generating curiosity, building confidence or creating a sense of pride are among the other options. In general, anything that affects “feelings” can be an emotional motivational tool.

Rational
Rational motivation focuses on "what's in it for me." It is an intellectual calculation involving netting attraction (benefits) and aversion (costs). The more that benefits exceed cost, the more rational motivation.

The costs of attending learning events are largely personal. The choice of the date, the location and current workload are examples of things that can affect aversion. These are all local to an individual. Things that can be done but generally aversion is difficult to influence.

Benefit is easier to effect. The first thing is to get them noticed. Showing that benefits are wide in scope, long in duration, high in certainty and quick to realize magnifies attraction. Table 7 gives a useful formula for maximizing the benefit component of rational motivation.

Table 7
PROCESS FOR INSURING BENEFIT VISIBILITY

1. State a feature of the learning. For example, “you will learn to direct potential conflict into productive channels.”
2. Assert an objective result. For example, “you will learn to reduce the stress on yourself and everyone around you.”
3. Specify an advantage arising from the result. For example, “you will develop a reputation as a person who can handle difficult situations and make a silk purse out of a sows ear.”
   
   

FITTING THE KOLB MODEL
A previous blog demonstrated that the Kolb Learning model is an applied subset of the “I Opt” information-processing model. “I Opt” adds exact measurement to the Kolb skeleton. This means that individual learning profiles can be specified. Examples of different learning profiles of actual people participating in this study are shown in Graphic 1.

Graphic 1
INDIVIDUAL LEARNING PROFILES

An "I Opt" learning profile defines “how” but not “how much” a person wants to learn. The measures outlined in this blog can be used to estimate the desire to learn. The 2D learning profile of Graphic 1 can be expanded to 3D to show this. The height of the profile describes the degree of motivation (see Graphic 2).

Graphic 2
3D LEARNING PROFILES

Individual motivation does not affect the teaching strategy. All of the learning quadrants are equally affected. However, when applied to groups, things change.

Graphic 3a (left) combines the learning profiles shown above. Overlaying the two-dimensional learning profiles defines the best way to teach. In this case, stressing the “I Opt” Changer (Kolb Accommodating) and “I Opt” Performer (Kolb Diverging) learning styles about equally will maximize total knowledge transfer.

Graphic 3
JOINT LEARNING PROFILES

Individual Graphic 3b adds motivation to the 2D graphic. The person favoring the "I Opt" Changer style is more motivated. Putting more stress on her favored Changer (Kolb Accommodating) style would result in more information being available to the firm. The Changer person would absorb more. The person favoring the "I Opt” Performer (Kolb Diverging) style would learn but would not be optimized.

This illustration shows a largely unrecognized trade off inherent in teaching. If you use learning styles and ignore motivation, you optimize the scope of knowledge transfer. You will distribute knowledge broadly. If you use motivation along with learning style, you optimize the amount (i.e., depth) of information transferred. The cost is that this knowledge may be resident in fewer people.

There is no “right” or “wrong” in the choice. There are situations where either strategy might be best. For example, you may be teaching a team who will work together regularly. In this case having the knowledge available in someone is what counts; its distribution is secondary.

If, however, you are teaching people from a variety of departments your goal is probably the dispersal of knowledge. In that case the 2D choice of using the learning styles alone would be best. Less information would be installed but it would be more widely dispersed.


MOTIVATING GROUPS WITH THE KOLB MODEL
What can be done with two people can also be done with larger groups. “I Opt” technology is also not confined to illustrations. Actual calculations are possible. Graphic 4 shows the technology applied to an actual class of 7 people in a Time Management class.

Graphic 4
APPLYING THE 3D MODEL TO GROUPS

Graphic 4 shows a 3D graphic of the 7 people with the percentage allocation of each teaching style indicated. This percentage was obtained by calculating the volume of each person's 3D learning profile in each quadrant. Adding the volume of the 7 profiles gives a measure of "motivated learning" preferences in each quadrant. Proportioning these preferences indicates how much attention should be given to each learning style.

The above calculations relaxed mathematical purity. The exact (i.e., ratio) scales of "I Opt" were multiplied by the ordinal (rank order) scales of motivation to get volume. This is a bit of a mathematical leap. However, the direction is probably right if not precisely accurate.


CONCLUSION
Motivation has long been recognized as a component of learning. This blog advances prior work by showing how it can be systematically managed. It then shows how it can be seamlessly integrated into the well-accepted Kolb learning model. In this process, a poorly recognized tradeoff between scope and depth of knowledge transferred has been highlighted. While not the final word on the subject, it is hoped that this will be a step forward.

Optimizing the Kolb Learning Model

by:
Gary J. Salton, Ph.D.

FORWARD

The Kolb learning model (Kolb & Fry, 1975) is the dominant paradigm in adult learning. It holds this position because it is a practical, easily understood taxonomy that works to an acceptable degree in field settings.

Kolb has its limits. Most importantly, it is based on psychology—the science of the individual. Most instruction occurs in a classroom setting. This is the realm of sociology—the science of groups. Kolb provides no set way of getting from the individual to the group. This is a serious shortfall for those interested in large scale training efforts.

This research blog estimates the size of the shortfall. Data was collected from 185 actual adult classes involving 3,116 individuals in 5 different organizations. Classes were held in cities in the United States, Asia and Europe. While not random, the sample is broad enough to suggest a level of credibility.

"I Opt" technology was applied to the data to bridge the psychology-sociology gap. "I Opt" is a theoretically founded, fully validated and tested knowledge base (see www.oeinstitute.org for "I Opt" theoretical underpinnings or www.iopt.com for operational tools). "I Opt" is an information-processing tool. Since information processing is common to both individuals and groups is a natural bridge between the individual and group.

ALIGNING THE KOLB MODEL

Both Kolb and "I Opt" use a four-axis model. Each axis of the "I Opt" model is a combination of input and output elections. Some of the behaviors produced by these interactions can be seen as the "orientations" that Kolb uses in his taxonomy (i.e. classification system). In other words, the Kolb conceptual model can be seen as a subset of the broader "I Opt" paradigm.

GRAPHIC 1 KOLB LEARNING MODEL AND "I OPT" GRID

Graphic 1 contrasts Kolb and "I Opt". Kolb is a conceptual scheme whereas "I Opt" is a measurement grid. The axes on the "I Opt" model are called "styles" whereas Kolb calls his conceptual variables "stages." The difference is more than just names. Kolb's model rests on MBTI psychology. There is no way to measure his stages except by asking people for their subjective assessment. There is no way to know if what one person means by "sometimes" is the same as what another person means. You can add Kolb's scores but you could be adding apples and oranges. You would never know.

The "I Opt" input and output elections are measured on an absolute rather than subjective scale. The absolute scale is what allows the elections of individuals to be combined into groups. This ability to "add up" people is what gives "I Opt" the power to bridge the individual-group gap.

"I Opt" can combine its style "building blocks" to produce different things (see here for list) . For example, the "I Opt" Hypothetical Analyzer style uses structured input and thought (conceptual) output. Structured input (a method that uses some form of reasoning to select input) combines with conceptual output (logical combination of particulars). This produces systems, theories and explanations. These products are all various levels of abstraction and are roughly consistent with Kolb's abstraction category. Kolb's other conceptual variables can be similarly derived from the "I Opt" model.

The match between "I Opt" and the Kolb is not perfect. However, it is sufficiently aligned to allow the data collected within an "I Opt" framework to be applied to the Kolb model.

MEASURING THE STRENGTH OF KOLB’S STAGES

Kolb sees a circular or spiral sequence of stages. Without measurement Kolb's model cannot tell you how much emphasis to give each stage. Rather, he suggests "touching all the bases." Without measurement there is no way of defining how hard the bases are to be "touched." Applying "I Opt" measurements from 3,116 people allows us to define what Kolb cannot.

GRAPHIC 2
KOLB’S “CONCRETE EXPERIENCE” AND “NEW SITUATIONS” VARIABLES

Percent of Students at Various Commitment Levels: "I Opt" RI and RS Styles

Graphic 2 reveals that 81% of people do not heavily rely on the Concrete (1) or New Situations (4) stages ("I Opt" RS and RI styles). Fully 48% of people prefer very low levels. Instruction that exceeds these levels will appear irrelevant and difficult to assimilate.

GRAPHIC 3
KOLB'S "OBSERVE & REFLECT" AND "CONCEPTS" VARIABLES

Percent of Students at Various Commitment Levels: "I Opt" LP and HA Styles

Graphic 3 shows that Kolb's Observe & Reflect (2) and Concept (3) stages operate in almost the reverse fashion. Fully 42% of people would prefer medium-high to very high levels of this instruction. Falling short of this standard will leave people feeling inadequately prepared.

Graphic 4 shows a cumulative distribution of the four Kolb stages. Extending a line form a desired coverage level shows the level emphasis needed. The line on the graph shows that targeting of 75% coverage would require different emphasis on the various Kolb stages.

GRAPHIC 4 CUMULATIVE DISTRIBUTION OF KOLB VARIABLES

"I Opt" is not confined to Kolb's ordinal (e.g., low, medium, high) measures. The actual strength of each style (Kolb stage) in Graphic 4 can be measured. Graphic 5 shows a pie chart distribution of these strengths. Because the I Opt scores are not simple categories but rather are the result of information processing elections, the pie chart can also tell you how to design instruction in terms of input-output structure.

GRAPHIC 5
MINIMUM LEVELS OF KOLB STAGES TO CAPTURE 75% OF STUDENTS

Graphic 5 shows that the designer should give about equal attention to thought and action—the "how" and "why" of the subject. However, in terms of input a 60-40 rule seems optimal. About 60% of instruction should be structured—a “this causes that” or “do this, then do that” sequencing. This is knowledge in a firm framework.

The remaining 40% should be unpatterned input. This is more spontaneous and less predictable knowledge flow. It typically involves decisions by students. For example, students might be asked to provide examples of applications or be asked to apply the principles to a subject. Their choice of examples or methods of applying principles will vary. This is knowledge in a "fuzzy" framework.

HOW TO MEASURE KOLB LEARNING STYLES IN GROUPS

Kolb recognized that people learn by combining his stages. He calls these combinations of stages “learning styles.” These are shown as diagonals in Kolb’s basic model. I Opt also has diagonal interconnections. I Opt uses the term “patterns” to describe these interactions. Graphic 6 compares the two approaches. The parallels are obvious.

GRAPHIC 6
KOLB LEARNING MODEL STYLES AND "I OPT" GRID PATTERNS

Visual similarity does not mean identity. “I Opt” technology has exact measurement. The Kolb learning styles can never be measured. This is because the scales used for each stage do not have a common denominator. "I Opt" is different. Each axis has an input or output election in common with an adjacent axes. This commonality along with identical axis scoring methods means that diagonals can be quantified. This means learning styles can be measured exactly.

Graphic 7 shows how measurement is done. The profiles of two real individuals from the 3,116-person research base are shown in Graphic 7A. Deciding how to teach these two people together would be a challenge if you were limited to the Kolb taxonomy.

GRAPHIC 7
MEASURING LEARNING STYLES OF GROUPS

“I Opt” offers a solution. "I Opt" profiles are not mere illustrations. They are exact measurements of information processing preferences. They can be superimposed. This creates a “common area” that describes an information flow acceptable to both people. The common area for Dave and Teresa is shown in white in Graphic 7B. The percentage of common (i.e., white) area in each quadrant describes the amount of that kind of instruction that will jointly serve the people involved. Clearly, touching all bases equally is a poor solution for this pair.

What can be done for 2 people can be done for an entire class. Graphic 8 shows the profile of the class in which Dave and Teresa participated. The gray area is the common area of a majority of the 16 people in the class (i.e., at least 9 people fall in every point in this area). This majority area is where most people share a common learning preference. If your purpose is to teach as many people as much as possible, this is majority area defines your optimal strategy.

GRAPHIC 8
LEARNING PROFILE OF AN ACTUAL CLASS

The yellow dots are in Graphic 8 are the centroids of each person in the class. A centroid is the single point that best represents a person’s whole profile (i.e., Cartesian Average). Centroids help give a quick picture of how individual people fit into the group as a whole. For example, this one shows that both Dave and Teresa are likely to be a bit dissatisfied with the group strategy. This is another advantage of “I Opt.” You can see who will need to be compromised and by how much in order to serve most people in the class. Taxonomies can never give this kind of information.

THE “TYPICAL” CLASS

Individual classes vary widely in their learning preference. Graphic 9 shows a random selection drawn from the 185 classes in which the 3,116 people used in this research participated.

GRAPHIC 9
KOLB LEARNING STYLE SAMPLE OF ACTUAL CLASSES
Actual classes conducted within the 185 class sample

All classes have at least some representation in each learning style (i.e., I Opt pattern). This means that almost any teaching strategy will work to some degree. However, the differences between classes illustrate that no instructional strategy that will optimize all of them. What ever you do will be a compromise. The issue is whether your compromise is optimal.

INSTRUCTIONAL DESIGN OPTIMIZATION
Instructional Designers are concerned with optimizing the learning of groups, not individuals. This means focusing on the learning styles (rather than stages) of the actual classes that will be taught. Graphic 10 shows the distribution for each of the four learning styles among of 185 classes measured in this research.

GRAPHIC 10
KOLB LEARNING STYLE DISTRIBUTIONS
Applied to the profiles of 185 classes

Graphic 10 also shows the midpoint of the each of the Kolb learning styles (i.e., "I Opt" Patterns). The midpoint (median) is where half of the classes fall on either side. A strategy of targeting the midpoint would minimize the degree that a learning style will exceed or fall short in a typical class. Instruction designed to hit this point would minimize the deviation an instructor would have to adjust for in particular classes. It is a viable option.

Graphic 11 compares the "touch all styles equally" and the midpoint design strategy. There is a 70% overlap in the two learning profiles. However, it also shows that designers should plan to give about twice more instruction in the Assimilating (B) than the Accommodating (D) learning styles. The difference between a 70% and 100% overlap can mean a lot.

GRAPHIC 11
“TOUCH ALL BASES EQUALLY” VERSUS MIDPOINT STRATEGY
Based on 185 Classes

This blog uses a large and diverse research population. However, the other blogs in this series show that there can be systematic differences by firm, function or organizational level. The safest course in designing instructional policy is to sample the actual population being taught and use it as a standard. In final analysis there is no substitute for exact measurement.

SUMMARY

The research has also shown that improvement in the Kolb model is possible. "I Opt" technology allows its concepts to be extended to groups. It can be used to measure the systematic biases in their learning preferences. It can show you exactly which individuals will be "short changed" when using an instructional design that is best for the group. Finally, it illustrates that reconciling psychology and sociology can magnify the reach and power of the learning professional.



BIBLIOGRAPHY
Kolb. D. A. and Fry, R. (1975) 'Toward an applied theory of experiential learning', in C. Cooper (ed.) Theories of Group Process, London: John Wiley.