Three Types of Learner Control

Although the term learner control is often used generically, the actual type of control varies. Thus, two courses that are depicted as “learner-controlled” may in fact offer quite different options. In general, control options fall into three domains:

1. Content Sequencing. Learners can control the order of the lessons, topics, and screens within a lesson. Many e-courses such as the design in Figure 15.1 allow content control through a course menu from which learners select topics in any sequence they wish. Likewise, links placed in lessons can lead to additional pages in the course or to alternative websites with related information.
2. Pacing. Learners can control the time spent on each lesson page. With the exception of short video or audio sequences, a standard adopted in virtually all asynchronous e-learning allows learners to progress through the training at their own rate, spending as much or as little time as they wish on any given screen, such as by including a “next” or “continue” button. Likewise, options to move backward or to exit are made available on every screen.
3. Access to Learning Support. Learners can select or bypass instructional components of lessons such as examples or practice exercises. Within a given lesson, navigation buttons, links, or tabs lead to course objectives, definitions, explanations, additional references, coaches, examples, help systems, or practice exercises. In contrast, a program-controlled lesson provides most of these instructional components by default as learners click the forward button.

Figure 15.2 shows a screen from an asynchronous course that allows control over all three of these domains. At the bottom right of the screen the directional arrows provide for movement forward or backward at the learner’s own pace. The course uses Microsoft standard control buttons in the upper right-hand corner of the screen as well as an on-screen button to exit. In the left-hand frame, the course map allows learners to select lessons in any sequence. Within the central lesson frame, the learner can decide to study the examples by clicking on the thumbnail sample screens to enlarge them. Learners can also select a practice exercise by either clicking on the link above the examples or on the navigational tab on the right-hand side. In addition, embedded links lead to definitions of terms. Table 15.1 summarizes the most common techniques used to implement various forms of learner control in asynchronous e-learning.

Tradeoffs to Learner Control

Advocates for learner control propose the following benefits. First, offering choices has positive motivational benefits, leading to persistence. Second, giving learners control actively engages them in the learning environment, leading to better learning outcomes. Third, making choices will help learners build self-regulatory skills that will pay off in better self-management in other domains.

In contrast, opponents suggest that high levels of learner control will result in extraneous cognitive load, which will waste valuable mental resources that could be devoted to learning. In addition, many learners lack the background knowledge or skills to make good decisions for themselves.

Rather than advocate for or against learner control, we provide guidelines and illustrations for when and how learner control is best used. Additionally, we describe some alternatives to learner control, including shared control, advisement, and recommender systems.

Calibration Accuracy: Do You Know What You Think You Know?

Suppose you have to take a test on basic statistics. Prior to taking the test, you are asked to estimate your level of confidence in your knowledge. You know that, even though you took statistics in college, you are a little rusty on some of the formulas, but you figure that you can score around 70 percent. After taking the test, you find your actual score is 55 percent. The correlation between your confidence estimate and your actual performance is called calibration. Had you guessed 55 percent, your calibration would have been perfect.

The focus of calibration measurement is not on what we actually know, but on the accuracy of what we think we know. If you don’t think you know much and in fact your test score is low, you have good calibration. Test your own calibration now by answering this question: What is the capital of Australia? As you state your answer, also estimate your confidence in your answer as high, medium, or low. You can check your calibration on the following page.

How Does Calibration Affect Learning?

Overestimates of your knowledge lead to overconfidence, with subsequent premature termination of study and practice. Dunlosky and Rawson (2012) manipulated learners’ accuracy judgments during practice exercises. Learners were presented with words and definitions for study. Next, they were given a word and asked to provide the corresponding definition, along with a self-assessment of the accuracy of their answers. After making their assessments, learners viewed a partial definition in the form of main idea units. Learners were required to continue to study all words until they rated their answers as high accuracy three times, at which point that word was dropped from the list (regardless of their actual accuracy). Levels of overconfidence were determined by the percentage of responses judged as correct that were actually incorrect. The research team then compared scores on the final test with the percentage of overconfidence during the study period.

The results showed that those with the most overconfidence scored lowest on the final test, while those who were most accurate in their self-assessments scored highest. For example, those whose overconfidence fell in the 50 to 100 percent range scored around 30 percent on the test. In contrast, those whose overconfidence estimates were 20 percent or less scored 80 percent and higher. The research team concludes that: “judgment accuracy matters a great deal for effective learning and durable retention: overconfidence led to the premature termination of studying some definitions and to lower levels of retention” (p. 7).

How Common Is Overconfidence Among Learners?

Although most of us feel we have a general sense of what we do and do not know, our specific calibration accuracy often tends to be poor (Stone, 2000). Glenberg, Sanocki, Epstein, and Morris (1987) found calibration correlations close to zero, concluding that “contrary to intuition, poor calibration of comprehension is the rule, rather than the exception” (p. 119). Eva, Cunnington, Reiter, Keane, and Norman (2004) report poor correlations between medical students’ estimates of their knowledge and their actual test scores. When comparing knowledge estimates among year 1, year 2, and year 3 medical students, there was no evidence that self-assessments improved with increasing seniority. The team concludes that “Self-assessment of performance remains a poor predictor of actual performance” (p. 222).

Now let’s check on your calibration. Review your response to our question on the previous page about the capital of Australia. The capital of Australia is not Sydney, as many people guess with high confidence. It is Canberra. If you guessed Sydney with low confidence or if you guessed Canberra with high confidence, your calibration is high!

In comparing calibration of individuals before and after taking a test, accuracy is generally better after responding to test questions than before. Therefore, providing questions in training should lead to more accurate self-assessments. Walczyk and Hall (1989) confirmed this relationship by comparing the calibration of learners who studied using four resources: text alone, text plus examples, text plus questions, and text plus examples and questions. Calibration was best among those who studied from the version with examples and questions.

Do Learners Like Instructional Methods That Lead to Learning?

Most courses ask learners to evaluate the quality of the course with an end-of-course rating sheet. Do you think there is a high relationship between these end-of-course learner ratings and actual learning? Sitzmann, Brown, Casper, Ely, and Zimmerman (2008) correlated approximately 11,000 student course ratings with after-training knowledge measures. The correlations were low at .12. Remember that correlations range from –1 to + 1, with values around 0 indicating no correspondence whatsoever among the variables. The research team concludes that “reactions have a predictive relationship with cognitive learning outcomes, but the relationship is not strong enough to suggest reactions should be used as an indicator of learning” (p. 289).

Consider an animated lesson for engineering trainees demonstrating a six-step procedure for maintaining a mechanical device, which shows each step as a simple animation initiated by clicking on a button (that is, low control). Do you think students would learn better if they were allowed to control the animation by rotating the objects through dragging motions and zooming through pinching motions on a touch screen with an iPad (that is, high control), as illustrated in Figure 15.3? In a recent set of experiments involving Brazilian engineering students, students liked the high-control version of the lesson much better than the low-control version, but did not learn significantly more (Pedra, Mayer, & Albertin, 2015).

Do students learn more when matched to their preferred instructional methods? Schnackenberg, Sullivan, Leader, and Jones (1998) surveyed participants before taking a course regarding their preferences for amount of practice—high or low. Participants were assigned to two e-learning courses—one with many practice exercises and a second identical course with half the amount of practice. Half the learners were matched to their preference and half mismatched. Regardless of their preference, those assigned to the version with more practice achieved significantly higher scores on the post-test than those taking the version with fewer practice exercises.

The bottom line: There is little correspondence between learner perceptions of lesson effectiveness and actual instructional value. In short, liking is not the same as learning.

Psychological Reasons for Poor Learner Choices

We’ve seen that calibration research as well as correlations between student ratings and student learning suggest frequent inaccuracy in assessing learning needs, with consequent overconfidence and poorer learning outcomes. Metacognition refers to learners’ awareness and control of their own learning processes, such as assessing how well they understand a lesson or knowing how best to study to achieve a learning goal. Metacognition is the mind’s operating system. In short, metacognition supports mental self-awareness and self-regulation. Individuals with high metacognitive skills set realistic learning goals and use effective study strategies. They have high levels of self-regulation skills. For example, when faced with a certification test, they plan a study schedule. Based on accurate self-assessments of their current strengths and weaknesses, they focus their time and efforts on the topics most needed for success. They use appropriate study techniques based on an accurate assessment of the certification requirements. In contrast, learners with poor metacognitive skills lack understanding of what they know and how they learn, which will lead to flawed decisions under high learner control.

Moos and Azevedo (2008) compared metacognitive activities among high and low prior knowledge learners as they researched a hypermedia resource on the circulatory system. After a pretest to evaluate knowledge levels, college students were allowed forty minutes to study the circulatory system from an online encyclopedia that included articles, video, figures, and other information. Students were asked to talk aloud while they studied, and their self-regulatory patterns were compared. Learners with high prior knowledge used more planning and monitoring processes as they reviewed the materials. In contrast, lower prior knowledge learners did little planning or monitoring but instead took notes. Because planning and monitoring require working memory capacity, it is likely that low prior knowledge learners lacked sufficient mental resource for self-regulatory activities. The research team recommends adding guidance to hypermedia environments that will be accessed by novice learners. For example, adding frequent questions with detailed feedback may give learners a more accurate view of their learning needs.

How can you best apply the evidence and the psychology behind learner control to your design of effective e-courses? In the remainder of this chapter, we discuss the following evidence-based guidelines for the best use of learner control to optimize learning:

• Principle 1: Give experienced learners control.
• Principle 2: Make important instructional events the default.
• Principle 3: Consider alternatives to learner control such as shared control, advisement, or recommender systems.
• Principle 4: Give pacing control to all learners.
• Principle 5: Offer navigational support in hypermedia environments.

When to Give Learner Control

A commonly agreed on exception to the negative effects of learner control involves learners with high prior knowledge. Evidence suggests that learners with knowledge relevant to the lesson domain will not be harmed by a high learner controlled environment (Patall, Cooper, & Robinson, 2008; Scheiter, 2014). Another exception involves giving content control when there are few logical dependencies among the lessons or topics. In those situations, the sequence in which instructional elements are accessed will not affect learning. A third exception involves scenario-based courses in which the learner should have options to make decisions—even incorrect decisions—to build critical thinking skills. For example, the automotive troubleshooting course in Figure 1.5 (page 17) allows learners to select various test equipment in any sequence. At the end of each lesson, learners can compare their selections with expert selections and in that manner can learn from their errors. In summary, learner control is shown to have greater benefits when:

• Learners have prior knowledge of the content and skills involved in the training.
• The instruction is a more advanced lesson in a course or a more advanced course in a curriculum.
• Topics and lessons are independent of one another so that the sequence does not affect learning.
• Choices among lesson elements are an essential design element to help learners build decision-making skills.
• The course is of low complexity.

Advisement

With advisement, after completing an exercise, the system offers suggestions to learners regarding what task they might select next. For example, if the learner successfully completes a moderately difficult task with moderate support, the system would suggest they next try a task at the same level of difficulty but with less support. Contrary to their expectations, Taminiau and colleagues (2013) found better learning from the group that did not receive advice. The research team suggests that, rather than give explicit advice, better results might come from advice about the process the learners should take based on their own ratings and results, allowing them to make their own decisions. As with shared control, we need more research on what kinds of advisement, if any, promote learning.

Recommender Systems

If you have shopped online you have encountered advisement systems, often in the form of user ratings such as stars and comments. Have you found these ratings helpful? How could a recommender system be productively applied to instructional products? Ghauth and Abdullah (2010) tested a recommender system in which only previous learners whose test scores exceeded 80 percent were allowed to give ratings. Their recommender system included a content filter to help learners identify appropriate instructional items from a large pool accompanied by “good learner” ratings of the different options. In comparing software engineering students who did or did not have access to the recommender system, they found better average learning outcomes from the group using the system.

One challenge is how to define “better learners.” Test scores may not align with better job performance. We will look for more research on the benefits of various types of recommender systems, which are demonstrated to serve as a valid form of advisement regarding instructional quality of a lesson or course.

Although all three of these learner control alternatives seem potentially useful, evidence has yet to confirm their effectiveness. We will need a larger body of research to make recommendations.

Use Headings and Introductory Statements

Content representations such as headings and introductory sentences improve memory and comprehension in traditional text documents. For example, Lorch, Lorch, Ritchey, McGovern, and Coleman (2001) asked readers to generate summaries of texts that included headings for half of the paragraphs. They found that the summaries included more content from paragraphs with headers and less from paragraphs lacking headers. Mayer (2005b) refers to headings as a form of signaling—providing cues concerning the important information in a lesson. We recommend that similar devices be used in e-learning programs. Screen headings, for example, might include the lesson title followed by the topic. On-screen text segments and visuals should likewise be signaled with brief descriptive labels similar to paper documents.

Use Links Sparingly in Lessons Intended for Novice Learners

Avoid using links that take the learner off the teaching screen as well as links leading to important instructional events. By definition, links signal to the user that the information is adjunct or peripheral to the main content of the site. Learners will bypass many links. Based on the research described previously, we discourage using links for access to essential skill-building elements such as worked examples or practice, especially with novice audiences.

Niederhauser, Reynolds, Salmen, and Skolmoski (2000) presented two related concepts in two separate lessons. In each lesson, links led learners to correlated information about the concept in the other lesson. For example, if reading about the benefits of concept A in Lesson 1, a link would lead to benefits of concept B in Lesson 2 for purposes of contrast. They found that nearly half the learners frequently made use of these links. The other half either never used the links or used them briefly before abandoning them in favor of a more linear progression whereby they moved through one lesson from start to finish before moving to the other. Contrary to the authors’ expectations, they found that extensive use of the links was negatively related to learning. They attribute their findings to adverse impact of hypertext navigation on cognitive load.

If, however, your materials do include links, Shapiro (2008) suggests adding annotations to the links that give novice learners a short preview of what is behind the link or to judiciously highlight links that are especially relevant to a specific learning goal.

Use Course and Site Maps

A course or site map is a type of menu or concept map that graphically represents the topics included in a course or reference resource. Nilsson and Mayer (2002) define a concept map as “a graphic representation of a hypertext document, in which the pages of the document are represented by visual objects and the links between pages are represented by lines or arrows connecting the visual objects” (p. 2). Figure 15.5 shows three different formats for course maps.

Research has been mixed on the contribution of course maps to learning. Niederhauser, Reynolds, Salmen, and Skolmoski (2000) included a topic map containing a graphic representation of the hierarchical structure of the hypertext. Learners could access any screen in the hypertext from the topic map. A trace of user paths found that many learners did access the topic map frequently but rarely used it to navigate. Most would access the map, review the levels, and return to where they were reading. A few participants never accessed the topic map. In correlating map use with learning, the research team found only a slight benefit. Potelle and Rouet (2003) compared comprehension of a hypertext between novice and content specialists for the three menu layouts shown in Figure 15.5: an alphabetical list, a hierarchical map, and a network map. Low knowledge participants learned most from the hierarchical map, whereas the type of map made no difference to high prior knowledge participants. It may be that course maps are less important for navigational control than for providing an orientation to the content structure—especially for novice learners.

We recommend the following guidelines regarding site maps:

• Consider using course maps or site maps for resources that are lengthy and complex and/or for learners who are novice to the content.
• Use a simple hierarchical structure.
• If your content will apply to learners with different tasks and instructional goals, consider multiple versions of a site map adapted to the instructional goals.

Provide Basic Navigation Options

In asynchronous e-learning, make elements for forward and backward movement, replay of audio and video, course exit, and menu reference easily accessible from every display. In courses that use scrolling pages, navigation should be accessible from both the top and bottom of the page to avoid overloading learners with unnecessary mouse work (having to scroll back to the top of the page to click “next”). Additionally, some sort of a progress indicator such as “Page 1 of 10” or a progress bar is useful to learners so that they know where they are in a topic and how far they have to go to complete it.

Chapter Reflection

1. Review a specific e-learning course and list the control options (learner or program) for the main course elements (pacing, content, instructional methods). For the intended audience, do you think the control decisions are appropriate?
2. What learner control standards has your organization established? If you were setting learner control standards, what would you define for (a) continuing nursing education or (b) introductory physiology for high school students?
3. Adult learners generally expect high levels of control on the Internet. How would you reconcile these expectations with instructional benefits of program control for novice learners?