Key Design Considerations for Personalized Learning on the Web
Margaret Martinez, Ph.D.
Designers are striving to build and support learning environments and solutions that encourage and enable learners to stay abreast and comfortable with new technology, constant change, and continual improvement. This article is about building interactive learning solutions that recognize, match, and support critical factors that influence how individuals learn. It introduces a whole-person approach focused on emotions and intentions to influence learning and performance improvement and offer guidelines for design and development.
To address today's sophisticated online learning and performance needs, the consideration of a comprehensive set of human factors that contribute to online solutions from a whole-person perspective is critical. Traditionally, designers have focused on primarily cognitive factors (e.g., how people build, process, and store knowledge). These primarily cognitive designs often overlook other sources for individual learning differences, such as emotions and intentions. If even they are considered, they are often relegated to a secondary role. Historically, these approaches reflect the industrial-age perspective that assumes that an instructor is available in the classroom and can respond to the audience's diverse range of complex human needs. As a result, instructors have often unintentionally created a learning dependence that detracts from the learner’s ability to self-motivate, self-manage, and self-assess online learning.
This paper is aimed at readers examining ways to design Web learning environments and solutions. The goal is to provide instruction that not only helps learners accomplish instructional objectives, but more importantly helps them become more successful, continual online learners. This paper suggests that conventional classroom design methodologies and solutions may need enhancement for online learning, and designers need alternative learning models that help learners become more independent, self-motivated, and self-managed online learners.
After reading this paper, the author hopes that readers will expand their understanding about learning by also considering (a) the influence of a comprehensive set of key psychological factors (conative, affective, social, and cognitive) that influence learning differently, (b) the often overlooked dominant impact of emotions and intentions on learning, (c) critical human relationships between learning environments, key psychological sources that influence learning differently, and learning ability, and (d) new ways to design supportive learning environments that individually adapt to how people want to learn, from the more comprehensive perspective.
As learners move online, how can we attend to the basic human attraction for individualized attention? How can we support more self-motivated, independent, and self-directed online learning? How do we provide designs that acknowledge learners as feeling, intentional, thinking, and social human beings, each self-managing a diverse set of personal traits that influence learning? How do we provide instruction and environments that match how individuals intend to learn differently, improve learning ability, and continually foster increasing expertise and satisfactory learning relationships?
Many theories and models exist today to address these complex issues. However, a model that is seldom used is one that recognizes the dominant impact of emotions and intentions on how humans learn. For example, it is important to consider how learners may want or intend learning differently, e.g., more passionately or less enjoyably. Other examples are one learner may want or intend to learn more independently or less independently than another or one learner may enjoy setting and accomplishing more difficult learning goals than others.
Contemporary neuroscience research describes the evolution of the amygdala (which processes and imprints emotional responses) and explains how emotions and passions influence, guide, and, at times, override our cognitive processes (LeDoux, 1996). For example, the amygdala is a part of the neural system that plays a role in the evaluation of the emotional significance of environmental stimuli. The amygdala influences what gets our attention and how learning changes the pattern of brain activity and develops new connections in his brain. In the child development area of research, Woodward, (1998) showed that humans are highly goal-oriented and already have intentions to guide learning and development at age six months. Other diverse disciplines that highlight the impact of emotions and intentions on learning and performance are psychology, sports psychology, music, marketing, and advertising.
Conventional primarily cognitive models may ignore or demote these powerful influences that serve as sources for individual learning differences. These models typically rely on instructor intervention to manage emotions, intentions, and social issues, in addition to the cognitive aspects. For online learners, the lack of an instructor is emerging as a problem simply because many learners are not very experienced at self-motivated or self-managed learning. The new learner-difference models should address this problem by (1) explaining how emotions, attitudes, beliefs, and intentions (in addition to the more commonly recognized cognitive and social factors) and their relationships influence, support, or undermine learning experiences, (2) considering the impact of these factors and explain how they may influence more successful online learning, and (3) providing guidelines for designing instruction and learning environments that help learners not only meet instructional objectives, but also improve online learning ability.
We are still very much in the experimental stage for creating learning and design models for Web learning environments. Much still needs to be understood about designing successful environments, both technically and pedagogically. In 1957, Cronbach challenged the field to find "for each individual the treatment [solution or environment] which he can most easily adapt." He suggested that consideration of the treatments and individual together would determine the best payoff because we "can expect some attributes of person to have strong interactions with treatment variables.”
What is the model for designing individualized solutions for the best payoff? After reviewing 355 research studies, to examine what works or does not work, Russell (1997) stated that combined research results are largely ambiguous and that “individual differences in learning dictate that technology will facilitate learning for some, but will probably inhibit learning for others, while the remainder experience no significant difference.” He adds that “when lumping all the students together into a fictional “mass” those who benefit from the technology are balanced by a like number who suffer; when combined with the 'no‑significant‑difference’ majority, the conglomerate yields the widely reported “no-significant-difference” results.
The confusing or inconsistent results arising from the research literature indicates something critical is missing from our cognitive-rich learning constructs and contemporary educational technology theories. From another perspective, Snow and Farr (1987) suggested that sound learning theories are missing and realistically require “a whole person view that integrates cognitive, conative, and affective aspects.” These researchers advised that educators cannot ignore or overlook these key psychological aspects because they interact in important, complex ways to support learning and performance outcomes. Otherwise, explanations about learning will be ambiguous and isolated from reality (1987).
Learning Orientations Model
This article describes a study that uses Learning Orientations (Martinez, 1999b), as the whole-person learning model, to investigate learning in adaptive environments. Guidelines for developing the adaptive environments recognize a dominant influence of emotions, intentions, and social factors on how individuals learn differently. In contrast to conventional perspectives, the learning orientation model assigns cognitive factors to a secondary (albeit, still very important) role. The profiles reflect how emotions and intentions guide, manage, or help us develop cognitive ability (such as, learning preferences, strategies, skills, and processes). Learning orientations are helpful because they describe key attributes of the learning audience, including their proclivity to learn more successfully.
Learning orientations, developed during previous research (Martinez, 1999b), represent how strategically individuals (aggregated by varying beliefs, emotions, intentions, and ability) plan and set goals, commit and expend effort, and then experience learning to attain goals (described in Table 1).
There are four learning orientations: Transforming, Performing, Conforming, and Resistant. These profiles provide three specific scales for measuring key learner-difference attributes: Aspects of Emotions and Intentions, Strategic Planning and Committed Learning Effort, and Learning Autonomy (shown as column headings in Table 1). Learning orientations have a variety of uses, including helping (1) differentiate the audience for research, (2) guide analysis and design of research, instruction, and environments, (3) tailor solutions that improve performance, learning ability, and online relationships, and (4) making the overall learning experience more satisfying and worthwhile.
Note: In determining orientation, we must allow for the possibility of “situational performance or resistance.” Learners may improve, perform, or resist in reaction to situations of positive or negative learning conditions.
Table 1.Four learning orientations organized by three critical learner-difference attributes.
Research Study: Designing Adaptive Environments using a Whole-Person Perspective
In a previous research study, (Martinez, 1999b) investigated the effects of using one learning environment that adapted and matched three learning orientations (i.e., Transforming, Performing, and Conforming learning orientations). The purpose was to test what effects the adapted presentations had on the learning experience. The learning orientation model (created and tested in previous studies) served as a foundation to measure, analyze, and explain the effects and interactions on multiple dependent variables over three time periods.
This mass customization approach highlighted emotions and intentions as a dominant influence on learning. Study results demonstrated that a learner experiences greater positive effects to the extent that the instruction and environment can appropriately match and support the individual's learning orientation. The study results also provided evidence for (1) specific factors that impact how individuals learn differently and (2) creating design guidelines for Web learning solutions.
The study provided a Discovering the World Wide Web Internet course to participants who were randomly assigned to one of three research groups (described in Table 2). Each research group received the instruction in a Web learning environment that matched one of the three learning orientations. Hence, each research group received a different presentation (described in Table 2).
Table 2. Description of the three learning environments for three research groups.
Web Learning Environments
An instructional and research model, called the System for Intentional Learning and Performance Assessment (SILPA) was used to create the three adapted presentations in one learning environment. This model (developed by the author in a previous study) was used to deliver the course to the research groups. The SILPA can presents three different environments that can match the learning orientations, foster improved learning ability, and replace traditional, “one-size-fits-all” solutions. The SILPA can diagnose the audience in advance of the instruction with the electronic version of the Learning Orientation Questionnaire (LOQ). The LOQ is a 25-item questionnaire that uses the three construct factors (Conative/Affective Aspects, Strategic Planning and Committed Learning Effort, and Learning Autonomy) to measure and identify learning orientation (Martinez, 1998).
The Intentional Learning Training (ILT) was presented to the learners in the Web Learning environment (GROUP EX1). The purpose of this intervention was to provide online learning tools and suggest to learners how they might improve their online ability with tools for more self-directed learning. The heart of the SILPA model is a learning management and assessment framework called the “iCenter.” The iCenter offered the learning resources to help learners examine the content of the course, set goals, reflect on presentation preferences, and review cumulative information about scores. These tools enabled learners to manage individual learning performance for the domain of expertise in an organized problem-solving structure integrated with dynamic practice and assessment activities. The “iMap” is also part of the iCenter. It is a learning progress map for self-monitoring performance.
To focus this investigation, the author chose five research questions to examine variance, effects, and interactions. One is highlighted in this paper
1. Do learning orientations influence group interactions (Control Groups EX1, CO1, and CO2)?
Intervention and Experimental Research Design
An experimental factorial research design was developed to conduct a multiple repeated measures univariate analysis of variance (ANOVA). It was helpful in analyzing the independent and interactive effects of two independent variables (learning orientation and intentional learning training) on four dependent variables (satisfaction, learning efficacy, intentional learning performance, and achievement).
Table 3. Repeated measures research design for three research groups
Note. The table shows three research groups with or without the Intentional Learning Training (independent variable 1) and iCenter resources: Group EX1 is the experimental group, and Groups CO1 and CO2 are the two control groups. The three orientation categories differentiate the subjects within the three research groups. (Resistant learners are not included in this study). The Repeated Y Measures for the four dependent variables appear in columns A1, A2, and A3.
Learning orientations are also used in the research design. The purpose of their use in this manner is to create a separate dimension to (1) guide development of the research design, environment and instructional treatment and (2) differentiate the learning audience before introducing the treatment and examining the results. In contrast to a “one-size-fits-all” approach, learning orientations add the more human dimension to the treatment and examination of multiple learning variables. This step is especially important because it distinguishes learners as individuals with predominant psychological characteristics in comparison to traditional considerations of learners as a uniform group of human beings with a homogenous set of emotions, intentions, beliefs, goal orientations, and values. The introduction of learning orientations and multiple variables is an effort to reflect a more realistic understanding of learners and learning experience. In this study, learners are not expected to want or intend to learn, set goals, and benefit alike from the same treatment.
Seventy‑one individuals (49 women and 22 men, mean age = 22) volunteered to take the Web course. All subjects, adults from local businesses, universities, and households had very limited or no Web experience. They also had a lot, some or no computer experience and showed a desire to learn how to use the Web.
Overall, the treatment was accomplished in three phases:
Analysis: Repeated Measure Univariate ANOVAs
Four sets of Y measures for the dependent variables were collected. The first set was collected during registration. The remaining data sets were collected during the course on three different occasions (i.e., repeated subintervals of the instructional cycle) between and among the three research groups. The author added parameters for learning orientation (treated as a continuous subject variable) to a modified mixed model with repeated measures. This model was used in the SAS system (PROC MIXED) to conduct a series of univariate analyses of variances. The ANOVAs were also modified to view and examine the data stratified by learning orientations. These analyses took a more specific look at how learning orientation within and between groups performed within the three adaptive presentations.
Finally, the author provided bivariate plots of orientation for each of the dependent variables. This is a method to examine the dependent variable effects by ILO and GROUP or ILO and TIME. This method uses the weights to plot the regression lines between X and Y (y = a + bx). These two types of plots show the group and time effects on the X-axis, respectively, as intentional learning increases. Only a highlight of the combined set of analyses appears in this paper.
The series of analyses were helpful in examining differences by groups and by learning orientation within the groups, between the groups, and over time. The ANOVA results exhibited statistically significant ILO, GROUP, and TIME effects and interactions on the dependent variables. The combined results indicated the likelihood that learners enjoyed greater success in learning environments that adapted and supported their individual learning orientation. In contrast, the learners adapted less positively in the unmatched environments that conflicted with their learning orientation.
The evidence suggests that learning orientations suggest useful way to (a) provide theoretical foundations using a comprehensive view of learning, (b) recognize dominant psychological factors, other than just cognitive aspects, that influence learning (c) stratify and analyze the audience—an important aspect of determining what works for the audience, and (d) guide design, development, implementation, analysis, and evaluation of solutions or treatments.
Multiple repeated measure ANOVA results for four dependent variables
The results showed statistically significant ILO (learning orientation), GROUP (EX1, CO1, and CO2), and TIME (three instructional units) main effects and interactions for the four dependent variables.
Table 4. Statistically significant ANOVA results.
Group Means and Standard Deviations by Time
To supplement the four main ANOVA analyses on the four dependent variables, additional analyses included an examination of group means and standard deviations by time and overall for each dependent variable. These results, organized into sections for three dependent variables, appear in Table 5. Additionally, Section 3 in Table 5 exhibits detailed information on achievement; this data shows how individuals (organized by learning orientations) achieved within the groups.
Overall, these results show that Group EX1, the intentional learning environment, had the highest overall group means for satisfaction, learning efficacy and intentional learning performance. Group EX1 also had the highest group achievement means for transforming learners. Interestingly, if you look at the overall group means for achievement (Table 5: Section 3), the results are very similar (M = .83, M = .85, and M = .84). As expected, each group mean averaged out to the group’s majority orientation (performing learners). Yet, if we look closely at the same data (stratified by learning orientation in Table 5: Section 3), the results show that each of the learning orientation groups performed highest in the matching learning environment (EX1: M = 94 for transforming learners, CO1: M = 91 for performing learners, and CO2: M = 87 for conforming learners).
Section 1. Means for Satisfaction Dependent Variable by GROUP and TIME. This table uses a 5-point Likert scale (5 = This lesson is very enjoyable for me, 1 = This lesson is very frustrating for me). The higher the rating, the greater the satisfaction with the course.
Section 2. Means for Learning Efficacy Dependent Variable by GROUP and TIME. This table uses a 5-point Likert scale (5 = Very Satisfied with my learning progress, 1 = Very Dissatisfied with my learning progress). The higher the rating, the greater the learning efficacy.
Section 3. Mean Percentage Correct for Achievement Dependent Variable by GROUP and TIME. This table shows the mean achievement scores (1.00 = High, 0 = Low) by GROUP and subgrouped by learning orientation.
Table 5.Means for four dependent variables by group
Additional Research Results
This research study and results also provided information about each of the dependent variables. Only a portion of these results are presented in this article. Results focused on the Satisfaction Dependent are included below to illustrate some of the additional research.
Set of Results for the Satisfaction Dependent Variable
The first data set described in Table 6 shows the means for the satisfaction (SAT) dependent variable. The satisfaction dependent variable means were consistently higher for the Experimental Group EX1 at every time point. It is interesting to note that all group means decreased during the second time period. Later, each of the group means increased during the final time period. In comparison to CO1 and CO2, (a) EX1 and CO1's decrease in means during Time 2 was less than CO2's and (b) EX1's and CO2's increase in means for Time 3 was greater than CO1's. The web learning environment and content seemed to influence satisfaction more than content difficulty. Despite the easy to hard content difficulty of the course, the subjects often verbally indicated that they enjoyed learning this content on the Web.
Table 6. Means for Satisfaction (SAT) Dependent Variable by GROUP and TIME
This table shows how subjects rated themselves across the duration of the course using a 5point Likert scale (5 = This lesson is very enjoyable for me, 1 = This lesson is very frustrating for me). The higher the rating, the greater the satisfaction with the course. Time 1, Time 2, and Time 3 were the first, second, and third part of the course and represented three instructional units.
Figure 1. Group means by time for the satisfaction dependent variable.
Overall, comparing the observations for the first and final time period of the course, the increase was higher for the Experimental Group EX1, equal for the Control Group CO1, and lower for the Control Group CO2. The Experimental Group EX1 had the highest overall increase in course satisfaction.
Bivariate plot of orientation and the satisfaction dependent variable.
The PROC REG command in the SAS system, using unstandardized regression weights for the predicted intercept and slope by GROUP, provided additional information about learning orientations within groups. To examine the effects for satisfaction by ILO and GROUP, the weights were used to plot the regression lines between X and Y using the linear equations formula, y = a + bx.
Figure 2. Linear equations for satisfaction by GROUP showing the regression of Y on X .
Figure 2 depicts the ILO * GROUP interaction, that is, the learning orientation influence on satisfaction, as the individual's learning intentionality increases or decreases in each GROUP. The higher the dependent variable satisfaction (Y axis) rating (1-5), the greater the satisfaction with the lessons in the instructional unit. Conversely, the lower the satisfaction score, the greater the frustration with the lessons in the instructional unit. The higher the learning orientation score (X axis), the higher the learning intentionality.
When interactions are significant, the lines are unparallel and demonstrate that a great effect on the dependent variable occurs with some degree of influence from the indicated variable. These results show nonsignificant GROUP effects a learning orientation level or score centered at 4.0. Noticeably, as intentionality increases, we see significant GROUP effects beginning at a learning orientation score of 5.0 and clearly occurring at a learning orientation score of 6.0 and above. These results may suggest that the restrictive instructional setting offered by the Control Group CO2 influences a significant amount of learning frustration as the learning orientation increases. In contrast, the other two research groups offer learning support that positively influences course satisfaction for the three learning orientations.
In this study, the author used learning orientations to (1) guide instructional and research design, (2) differentiate the audience, and (3) investigate how individuals learn differently in the three adaptive environments (which either matched or mismatched the diagnosed learning orientation). The results added to an alternative understanding about the relationship between instruction, environments, learning ability, and a comprehensive set of higher-order psychological factors that influence learning. Finally, the results suggest that the use of new technologies on the Web is enabling adaptive instructional presentations and environments that can more closely match the learner’s appreciation of individualized instruction.
A discussion of the results appears in the context of the following research question:
Research question 1: Do learning orientations influence group interactions (Groups EX1, CO1, and CO2)?
The ANOVA results in Table 4 showed statistically significant ILO * GROUP interactions for satisfaction (F = 6.48, p < 0.01) and learning efficacy (F = 3.93, p < 0.05). These findings indicate how likely the interactions between learning orientation and group seem to have impacted satisfaction (99%) and learning efficacy (95%). These findings suggest that learning environments influence learning outcomes depending on how it matches the learning orientation. These results echo Cronbach and Snow’s evidence that continued to show that the learning outcomes were better when the instructor's presentation adapted to the student's aptitude and personality (1977). For example, the "constructively motivated student who seeks challenges and takes responsibility is at his best when an instructor challenges him and then leaves him to pursue his own thoughts projects" (Cronbach & Snow, 1977). Other analyses showed that how time was managed in these environments is also a relevant learning factor since the TIME effects were statistically significant for satisfaction, learning efficacy, and learning performance.
A comparison of the group means by learning orientation for achievement (Table 5: Section 4) showed that individuals did best in the environments which best suited their learning orientation. Other analyses supported this evidence as it showed how learners with higher orientations had higher achievement and improved learning performance in the more sophisticated learning environments.
The evidence suggested that recognizing a more comprehensive set of common learner attributes, such as those influenced by emotions and intentions, is useful in guiding the design of instructional solutions and environments that enhance the overall learning experience. However, it is also important to note that although learners did best in the environment that best suited their learning orientation, they were not in an environment that would help them experiment and improve online learning ability.
Similarly, Cronbach and Snow (1977) suggested that "in dividing pupils between college preparatory and non-college studies, for example, a general intelligence test is probably the wrong thing to use. This test being general, predicts success in all subjects, therefore tends to have little interaction with treatment, and if so is not the best guide to differential treatment. We require a measure of aptitude that remains to be discovered. Ultimately we should design treatments, not to fit the average person, but to fit groups of students with particular aptitude patterns. Conversely, we should seek out the aptitudes which correspond to (interact with) modifiable aspects of the treatment."
I will use these research findings to refine guidelines that are more reflective of differential treatments that are particularly sensitive to performing and conforming learner basic needs, yet can still promote improved learning ability, (e.g., self-managed, self-motivated learning). These development efforts will focus on making these learning orientations more comfortable, engaged, and willing to set higher standards and perform in the intentional learning type of environment (Group EX1). The goal is to help learners internalize new skills as they improve learning ability over time.
The study proposed a theoretical foundation for adaptive learning--from a whole-person perspective--that recognizes the special impact of emotions and intentions on learning. It discussed individual learning differences, personalized learning (differential treatments), and adaptive learning technology. Finally, this study ends with two suggestions. The first is that supporting individualized learning with a whole-person theoretical foundation is an important consideration for a more complete solution. It is particularly highlights this importance for online learners who need to become more self-directed, self-motivated, and self-assessed. The second suggestion is that new instructional design and learning models should (1) reveal the special primary and secondary relationships between a more comprehensive set of psychological factors (conative, affective, social, and cognitive factors), (2) explain influences on the critical performance and achievement attributes that lead to more successful or less successful learning, (3) support differences in how people want and intend to learn, and (4) introduce new strategies that lead to improved online learning ability.
Web Design Principles Using the Whole Person Perspective
The study results provided suggestions for designing learning environments:
Since the study’s completion, other design guidelines have been added (shown in Table 6). These descriptions (organized by three learning orientations) are intended as general guidance for designing Web learning environments and instruction. They consider key issues that influence Web learning and provide information for accommodating the differences. Their overall purpose is to match the orientation to foster self-motivation, interest, meaningful interaction and more successful, independent learning. These same descriptions are also useful for creating a set of evaluation criteria against which Web instruction may be evaluated.
Table 6. Strategies and guidelines for three learning orientations
The statistically significant findings indicate that using a comprehensive set of factors, more than just the cognitive factors, is a step towards identifying the key sources for individual learning differences. It is also a step towards understanding and managing the impact solutions and environments have on different online learners. Also, suggested is that developing new learning models that highlight these factors is a useful way to understand the dominant (or most common) learning audience profiles before determining, designing, matching, and evaluating solutions and environments for more successful online learning.
Additionally, the evidence demonstrates the importance of differentiating the audience according to common attributes that meaningfully impact the proposed hypotheses and research results. In this study, the identification of learning orientations in the research design reflects that not all individuals have similar emotions and intentions to learn. The added dimension recognizes a richer palette of human interaction, including a whole-person emphasis considering emotions, intentions, and social factors, in addition to cognitive factors. This is an alternative perspective influencing research design, in contrast to the “one-size-fits-all” approach that may lead to Russell’s notion of “no-significant-difference” (1997). Additionally, this perspective considers the use of new technologies and explores how change in adaptive learning technology and infrastructure can accommodate individual learning differences successfully.
Hopefully, these suggestions will contribute to more successful learning via the Web with sound theoretical foundations and a greater understanding about fundamental attributes that create and support learning differences. When we design a course with only a universal type of learner in mind we unintentionally short hand learners or set them up for frustration and possible failure. If we are serious about providing good online instruction for learners, we should use the technology to provide instruction and environments in multiple ways so that all learners have opportunities for success, measured not only by meeting instructional and performance objectives but also by improved learning ability.