Cognitive-based methods to facilitate learning of software applications via E-learning systems

Abstract E-learning systems, which are used for teaching complex software, can facilitate learning if they provide an appropriate teaching approach that decreases learners’ cognitive load in addition to providing practical knowledge. We believe there is lack of cognitively guided educational recommendations on how to effectively and efficiently design such learning platforms. We thus provide an integrative review paper that overviews relevant literature to cognitive load theory to provide practical solutions and an empirically validated framework to decrease learners’ cognitive load and improve the learning of complex software through E-learning systems. The solutions (which contain practical examples) are proposed based on different concepts of cognitive load theory including using analogies, worked examples and infographics to facilitate schema acquisition; keeping learners’ concentration on the target tools by preventing split-attention and redundancy effects and applying the training wheel method; using interactive videos based on embodied cognition theory and finally considering the modality and transient information effects in designing E-learning systems. These solutions are related to adapting the learning platform to human cognitive structures and can lead to increased learning performance by preventing working memory from being overwhelmed, thus facilitating the formation of schemas and resulting in more efficient and reliable learning with less effort.


Introduction
In the past few decades that software applications have been taught with the use of the Elearning platforms, there has been a lack of specific guidelines for educational designers as to how they can design efficient learning platforms for teaching software applications. In order to use software applications efficiently, effective training methods that integrate psychologically based guidelines with modern training technologies are indispensable. Cognitive load theory (CLT) is a current educational psychological theory based on empirical evidence whose prescriptions and implications can have significant positive effects on learning computer science and related content and can provide the guidelines needed in this domain (Sweller et al., 2011a(Sweller et al., , 2019. This theory focuses on the characteristics of human cognitive architecture which refers to the elements of cognition including working memory, long-term memory and the way that they are organized and interact with each other (Sweller, 2006;Sweller et al., 2011a). Although, long-term memory capacity is not limited and humans appear to be able to memorize and recall unlimited information during their lifetime, working memory which is responsible for learning new concepts is limited and can be overwhelmed if the input information is out of its capacity (Cowan, 2001). Cognitive load is defined as the total amount of mental effort which is utilized in working memory to learn a concept or solve a problem (Sweller, 1988). CLT places emphasis on the limitations of working memory to accept new information during learning (Paas et al., 2004;Sweller, 1988;Sweller et al., 2011a).
Learning new software applications can involve high cognitive load as they often involve menus, tabs, and toolbars laid out with many different sections and a great number of tools and functions, that can make the application incomprehensible for novice users (Fang et al., 2011;Reis et al., 2012). Therefore, it is prudent to apply solutions that can decrease cognitive load and facilitate learning of software applications for novice users. In this paper, we conducted an integrative literature review on the principles of CLT, to propose a number of different ways of decreasing cognitive load in order to teach software applications to novices through E-learning systems, and we also reference associated empirical evidence. The purpose of an integrative literature review is to "provide fresh, new perspectives on the topic" and aims to generate "new knowledge about the topic reviewed" in an "integrated way such that new frameworks and perspectives on the topic are generated" (Torraco, 2016). We have fulfilled this purpose by looking at the learning of software applications for novice users from a novel perspective and examine the principles of CLT and their effects on the process of learning.
Although, E-learning systems are the most popular way of learning software applications, they cannot be considered an efficient approach if they are unable to regulate learners' cognitive load (Anderson, 2008;Mayer, 2005). In the next section different types of cognitive load will be explained, then different solutions to decrease cognitive load in the context of learning software applications will be referenced and finally a conclusion will be drawn.

Types of Cognitive load
There are three types of cognitive load including Intrinsic load, Extraneous load and Germane load. Each of these interact with and impact the total cognitive load of a learner. Intrinsic load is defined as the difficulty level of a specific instructional topic. It is related to the nature of a learning topic and is determined by the level of element interactivity of learning materials (Sweller, 1994). Element interactivity is an index to measure the complexity of a learning topic and depends on the learners' prior knowledge, nature of the materials and the relationship between the concepts that learners should connect and process simultaneously (Chen et al., 2015;Marcus et al., 1996;Sweller, 2010Sweller, , 1994. Elements with low interactivity can be learnt without or with minimal reference to other elements, for example, learning the features like bold, italic and underline in MS Excel. However, elements with high interactivity consist of different integrated elements that cannot be learnt in isolation, for example, typing functions in MS Excel (Sweller, 2010).
Extraneous load is defined as the difficulty level due to the way that information is presented. In comparison with intrinsic load that depends on the nature of instructional content, extraneous load depends on the way that instructional materials are designed and presented (Chandler & Sweller, 1991). Therefore, instructional designers can enhance learning performance by designing appropriate teaching materials that increase working memory resources by reducing extraneous load (J. Van Merriënboer & Sweller, 2005).
In contrast with intrinsic and extraneous cognitive load which emphasize the characteristics of the learning material, germane cognitive load is related to the learner's characteristics. It refers to the amount of working memory which learners allocate to dealing with the intrinsic cognitive load of teaching materials. If germane cognitive load is high, learning performance will increase, since germane load increases when extraneous load is low. Increasing germane load can help learners to devote a larger proportion of working memory resources to dealing with the target learning materials (Sweller, 2010). In other words, an instructional design that uses working memory capacity efficiently will decrease extraneous load, consequently germane load can be increased, and the formation of schema will be facilitated (Paas & Van Merriënboer, 1994).
A schema refers to the information that is stored in long term-memory as structured knowledge that helps problem solvers to recognize the type and category of a problem and retrieve an appropriate action to solve the problem (Sweller, 1988(Sweller, , 1994. In other words, schemas describe patterns of thought or behavior that are formed in the long-term memory based on previous experiences. Each schema is used to retrieve information for problems that completely or partially match that schema (Ericsson & Kintsch, 1995;Sweller, 1994). The process of schema acquisition is essential to learning and a long-term goal of learning is to facilitate this process and ultimately allow learners to automate schemas, which frees up working memory resources for other tasks (Sweller et al., 2011a).

Methods to facilitate schema formation in learners' long-term memory
Professional software users know where to find information, how to work with all the software's tools and have complete familiarity with the software layout, as well as, knowing how to perform different tasks using a combination of tools. These users have acquired complete schemas of the software in their long-term memory including software layout, placement of different tools and relationships among them. In order to facilitate the formation of schemas for novice users, we suggest the use of four specific instructional strategies including chunking contents, using analogies, infographics, and practical examples.

Chunking contents into the smallest components for high element interactivity tools
In order to use high element interactivity tools, users need to follow different connected steps in addition to having pre-required fundamental knowledge (Chen et al., 2017(Chen et al., , 2015Kalyuga, 2015). Although, there is no solution to decrease the element interactivity and intrinsic load of high element interactivity tools, we can decrease extraneous load and facilitate formation of a schema by providing necessary fundamental knowledge before teaching the tools and by breaking down instructions into the smallest teaching chunks such as step by step instructions instead of a long description (Sweller et al., 2011a;Van Merriënboer, 1997). In fact, isolated information can help novice learners learn each element of a complex concept separately without overwhelming their working memory which is limited to process 4 ± 1 unfamiliar elements at any given time (Cowan, 2001).
For example, typing a function-based formula in Microsoft Excel can be considered as a high element interactivity task, since users cannot use any function without having the necessary prerequired knowledge and without processing all the integrated steps of writing a function. Before choosing any function, users need to know the usage of each function. Then they should arrange different parameters in the function sections in the correct order. In order to teach how to write a specific function, first the usage of that function should be taught, followed by highlighting its differences to similar functions, then step by step instructions of the parameters that should be added in each section of the function should be communicated.

Analogies
One of the best solutions that can facilitate formation of schemas is linking the new concepts to learners' prior knowledge, by using analogies. Linking the layout and the tools of the new software to software that learners are already familiar with can facilitate learning through supporting the formation of schemas and consequently facilitating the acquisition of new knowledge (Dahl et al., 2008;Gick & Holyoak, 1987;Quiroga et al., 2004;Sweller, 1994). For instance, as it can be seen in Figure 1, in order to teach a graphical software tool such as Photoshop, if learners are familiar with a simple software product like Microsoft paint, we can teach some tool component in Photoshop by comparing it to its similar but less complex tool components in Microsoft paint and discuss the similarities and differences. In this case learners can connect the usage of the new tools to the tools that they know, and they can develop a deeper understanding of what they can do when using the new tools.  (Chen et al., 2015;Renkl, 2014, Sweller & Cooper, 1985. Worked examples can facilitate initial acquisition of cognitive skills and learning complex problem-solving skills through introducing step by step solutions for a formulated problem (Renkl, 2005). The main function of worked examples is to facilitate the construction of correct schemas for the learners, to then enable them to solve the relevant problems (J. Van Merriënboer & Sweller, 2005).
Worked examples are an effective method of learning, since they provide expert schemas to teach the procedure of a solution to a problem to novice learners. Based on Renkl (2005), in order to increase the efficiency of worked examples, the examples should be: a) self-explanatory, b) provide clear guidelines for the problem, c) show relations between different elements of the problem, and d) highlight main features in order to select the correct solution for the problem. Also, the effectiveness of a worked example depends on the element interactivity or complexity of the problem and its solution and the associated level of intrinsic cognitive load. Worked examples tend to be most effective when element interactivity is high, since when intrinsic cognitive load is high, the total cognitive load may exceed working memory capacity and controlling extraneous Linking the new concepts to learners' prior knowledge by connecting the tools of simple software that users are already familiar with to the tools of complex software that they are learning cognitive load becomes imperative (Chen et al., 2015). Therefore, applying worked examples could be very effective when it is used for teaching complex software tools.
Since each software tool is designed to facilitate the work that we perform manually in the real world, we can teach the software tools by using real world examples and problems instead of just introducing the tools' usages. In this case, the learners can get practical knowledge and extraneous cognitive load will decrease especially when the target tool is difficult to learn (Clark & Mayer, 2016;Sweller & Cooper, 1985). For instance, the "Patch" tool, in Photoshop software, that is used to remove some types of skin imperfections can be considered a high element interactivity tool for novice users as they need to follow a series of integrated steps to work with this tool. Therefore, applying a worked example could facilitate the learning of how to use the patch tool by showing learners the types of skin imperfections that this tool can remove and how this tool should be used. In order to design workexamples for the Patch tool a picture of a face with the type of skin imperfections that is compatible with this tool can be provided while showing step by step instructions of how the Patch tool can be used to remove imperfections from the face (see, Figure 2).

Narrative.
Narrative is a storyline and defined as an imaginary project or a scenario that is used to integrate the worked examples (Palomino et al., 2019). When using narrative to teach a software application, an imaginary real-world project is defined and in each lesson the tools necessary to complete a specific part of the project and their relationship with the previous tools are taught. This integration of the lessons within a familiar context can help decrease extraneous cognitive load and increase germane load by helping learners get an overview of the software tools and understand the relationship between them (Pollock et al., 2002;Van Mierlo et al., 2012). Since each software is a package of tools that users need to know the usage of each, in addition to how they relate to each other, therefore, using narrative can help provide a complete schema (Reedy, 2015).
There are different studies that have attested to the positive effects of narrative on learning. The best examples are Ribbon Hero that was developed by Microsoft to help users to learn useful features in MS Office (Shane, 2013) and GamiCAD to help new users learn AutoCAD (Li et al., 2012).

Figure 2. Example of worked examples.
In a large study by Darejeh et al. (2021), they showed that adding a familiar narrative into the Elearning system for teaching complex software, can increase learning performance significantly by decreasing cognitive load and providing practical knowledge to the learners.

Considering Expertise reversal and Guidance fading effects when using worked examples.
Although, presenting learning content in detail and applying different worked examples can be beneficial for inexperienced learners, by increasing the expertise level of the learners, worked example effectiveness can be decreased, eliminated or even reversed. This is referred to as the expertise reversal effect, where an instructional method that is best for inexperienced learners, may be ineffective for expert learners. This effect exists because essential information for the novices may become redundant for the expert learners and consequently may increase extraneous cognitive load (Kalyuga et al., 2003).
Therefore, after learners become familiar with the software layout and tools, in order to teach the relationship between tools and how combinations of different tools can be used, instead of providing a long description or different worked examples we can use problem-solving exercises with minimal guidance. This method can decrease extraneous cognitive load as we adapt the amount of guidance with learners' knowledge. This method is supported based on the guidance fading effect where instructors replace worked examples with completion problems thereby further increasing learners' knowledge (Kalyuga & Renkl, 2010;Nievelstein et al., 2013;Renkl et al., 2004;Stark, 1998;Van Merriënboer et al., 2002).
An appropriate completion problems exercise should lead users to work with the group of related tools and show them some hints while solving the exercise. For example, in order to teach how to have a professional face polish in photoshop, after users have learnt all the related tools for face polishing, a picture of a face with different imperfections can be provided to learners and they could be asked to use different face tools to polish the skin. In order to guide learners on how they should polish each imperfection, by clicking on each part of the face a hint box can open and shows learners what tool they should use (see, Figure 3).

Using Infographics/diagrams to improve recall and learning
After teaching all the steps of working with the target tool of the software, at the end of each lesson an infographic can be provided to summarize the instruction in a succinct and visual format, by indicating different steps of working with the target tool of the software visually (Lyra et al., 2016;Yildirim, 2016). These infographics are potentially useful as they summarize complex information which can lead to an improvement in recall and learning (Clarke et al., 2006;Gobert & Clement, 1999;Marcus et al., 1996).
Studies showed that presenting some parts of the content using an illustration can decrease cognitive load, however, the illustration should be simple and self-explanatory and there shouldn't be any text that learners need to integrate with the illustration (Chandler & Sweller, 1991;Purnell et al., 1991). Two types of infographics can be designed based on the number of steps that are needed to complete a task using software: 1) When teaching short tasks, just one screen of the software can be used to present all the steps needed to perform the target task. As an instance, Figure 4 shows an infographic for how to calculate summation in MS Excel: 2) When teaching long tasks, a sequence diagram can be used to show each group of steps on different software screens. The series of shorter but related diagrams help to reduce visual search and ensure the diagrams are less complex. As an instance, Figure 5 shows an infographic teaching how to add data to Excel from a text file:

Replacing static pictures with animation
Animation can help us to learn by observing instead of reading and hence omit the need to mentally connect text with pictures (Ayres et al., 2009). Animation effects occur when a series of static graphics that are used to teach a concept are replaced by an animation. Studies show that animation is superior to statics in learning procedural-motor knowledge where learners need to engage in a process using their body movements. In particular, animations are most useful for human movement related tasks (Van Gog et al., 2009). For example, the efficiency of animation is demonstrated in learning how to tie knots, use Lego to build different shapes, make origami shapes, and learning surgical skills (Ayres et al., 2009;Castro-Alonso et al., 2015;De Koning et al., 2019;Marcus et al., 2013;Masters et al., 2008). Therefore, one of the important factors that can affect the efficiency of animation in teaching software applications is the type of software and the content that is taught to the learners. Animation can be more efficient for teaching graphical software applications such as Photoshop or AutoCAD where learners need to learn drawing skills, however, it can be less efficient or even harmful for teaching the coding parts of the software where learners need to read and understand the code. We argue that applying signaling and using interactive animations are techniques that can make animations more efficient.

Signalling
Signalling can increase the efficiency of the animations by guiding learners' attention to important parts of the content through the use of different approaches such as highlighting, drawing a line, changing the color or adding an arrow to bold the target teaching content (Horvath, 2014;Mayer & Fiorella, 2014;Van Gog, 2014). In the context of teaching software, signalling can decrease extraneous cognitive load by drawing learners' attention to the target tool or setting, giving the learner the opportunity to ignore the irrelevant parts of the interface (Alpizar et al., 2020).

Interactive animations
Interactive animations enable learners to interact with the video content by moving their hand and clicking on the screen, which could have a positive effect on learning through simulating the software environment and engaging the motor system in learning. Darejeh (2021) showed that mouse movement and clicking on the video content can facilitate learning software applications, due to embodied cognition effects (Barsalou, 2008).
Embodied cognition theory emphasizes the role of the entire body including the motor system to achieve cognitive skills. This theory explains new ways of conceptualization in which human cognition can be affected by body movements and interacting with environmental objects (Shapiro, 2019;Wilson & Foglia, 2017) and the use of action to support educational goals (Weisberg et al., 2017). Many studies have proved that involving learners physically with the objects that are related to the learning content can enhance learning performance as there is a tight connection between cognitive processing and physical motion areas of our brain (Ceciliani, 2018;A Glenberg, 2015;Shapiro & Stolz, 2019). Therefore, integration of physical manipulation and imagined manipulation can enhance learning (A. M. Glenberg, 2010) by decreasing the cognitive load for difficult cognitive tasks (Calvo & Gomila, 2008;Schulz, 2017).

Figure 5. Example of an infographic for adding data to Excel from a text file.
Interactive animations can help users to learn how to use the tools of graphical software applications such as how to draw, how to change the brush size, how to use the stamp, or patch tools for repairing a photo. In order to use these tools users need to utilize a combination of keyboard keys and mouse movement, and so interacting with the software interface can increase learning performance by giving users practical tool usage learning activities that involve both mouse and keyboard manipulation. Moreover, this interaction can help learners to more easily remember the location of different tools within an interface, as instead of just watching an animation, they watch and click on software tools which can increase learning performance based on embodied cognition effects, where mind body interactions can be used to support cognition (Shapiro, 2019;Weisberg et al., 2017;Wilson & Foglia, 2017).
For example, in order to teach the rectangle tool in Photoshop using an interactive animation, the animation can be paused after explaining the usage of the rectangle tool and a notification can pop up on the rectangle tool and ask learners to click the tool to continue the video. After learners locate and then click on the rectangle tool, the video will continue automatically. Figure 6 shows an example of an interactive animation.

Keeping learners' concentration on the target tools
Since software applications have a crowded interface, it is important to keep learners' focus on the tools that are being taught. It can be done by considering split-attention effect, redundancy effect and applying the training wheel technique.

Avoiding Split-attention effect
Split-attention effect occurs when learners need to mentally integrate different information sources that are physically separate, such as text and pictures, into one schema in order to learn a concept. If learners cannot connect these sources their attention shifts since they need to keep information from one source that is active in their working-memory in order to understand the information coming from other sources (Ayres & Sweller, 2014). Chandler and Sweller (1992) showed that learning through integrated instructions can decrease learning time and increase test performance marks. Therefore, an efficient instructional design should try to integrate related graphical and textual information by placing them close to each other and in an appropriate place (Al-Shehri & Gitsaki, 2010;Schroeder & Cenkci, 2018).
The crowded interface of complex software can cause split-attention when learning through Elearning systems. In particular, while learning a specific tool that needs step-by-step instructions, the lack of integration of the instructions with different parts of the software can cause splitattention and decrease learning performance. This is because learners should read the instructional text and at the same time integrate the text with the software interface. Therefore, integrating the instruction steps with the software interface and locating the steps on the related part of the software interface, can prevent split-attention.
For example, in order to use the Stamp tool in Photoshop, users need to follow four steps. Reading these four steps and matching the instructions with the relevant component of the software interface would be more cognitively demanding than reading the instructions of each step within the relevant part of the software interface.
Therefore, the way that the stamp instruction is presented in Figure 7(a) can cause splitattention as the instructional text, and the software interface are not integrated and users should mentally integrate the step number and the instructional text with each other. However, as it can be seen in Figure 7(b), the instructional text is integrated into the software interface and learners do not need to match each instructional step with the software interface. Therefore, their concentration will be kept on just learning the software. This can prevent learners from being unable to locate the target teaching concept and their attention will not be wasted on the mental integration of the description text to its related tool (Ward & Sweller, 1990).

Avoiding the use of redundant and decorative elements
When a software tool is taught by using redundant pictures and auditory presentations, it can lead to lower learning results. This negative effect is referred to as the redundancy effect where cognitive resources are required to process the extra unnecessary information (Jin, 2012;Mayer, 2005). Since software has a graphical interface, we should avoid using extra pictures for explaining how to work with different tools. However, there can be exceptions for teaching critical concepts such as codes and syntaxis that learners need to memorize in software applications such as Microsoft Excel or MathLab. We can present the codes in the form of text in addition to a picture in the form of a diagram that explains the hierarchical structure of the code.
The decorative elements can distract learners and decrease their concentration level on the main learning contents (Craik, 2014). For example, Darejeh (2021) showed that talking avatars can decrease learning performance and increase learning time, due to the redundancy effect and an associated increase in cognitive load. Also, Goodell et al. (2006) demonstrated that decorative visual elements in a virtual reality surgical environment can decrease test performance marks.
The detrimental effect of the decorative elements can be increased when they are attractive for the users and can change the focus of their concentration (Mayer et al., 2008;Sung & Mayer, 2012). However, if the decorative pictures such as the instructor picture, or talking head play the role of a motivator, they can be harmless, but it depends on the type of learners and the difficulty level of the teaching contents (Lenzner et al., 2013;Morrison et al., 2001;Park et al., 2015;Schneider et al., 2018).

Training wheels
The final method that can keep learners' concentration on the tools that they are learning, is applying a training wheel. In this method, learners only see the settings of the target tool they have learnt in the previous steps and the current settings that they are learning (Bannert, 2000). The other tools and settings are hidden or greyed out to decrease cognitive load by eliminating the number of unfamiliar elements on the screen (Fang et al., 2011;Pociask & Morrison, 2004;Reis et al., 2012).

Figure 7. (a): Example of classic split-attention (b): Example of how to avoid split-attention.
The idea of training wheels is to introduce full functionality once the basics have been taught. They have the potential to reduce the cognitive load of novice users as there is less information and functionality for the users to consider and choose from. Training wheels should be used with care so users understand what full functionality will be made available to them down the line. For example, in order to teach how to enter a formula in Microsoft Excel to novice learners, the first step is teaching worksheets and how to enter data and function in the cells. Therefore, at the beginning of learning all the tools in the font section can be deactivated and greyed out to keep learners' concentration on the worksheet and data entry process (see, Figure 8).

Considering Modality and Transient information effects
Working memory has separate channels for processing visual and auditory information (Mousavi et al., 1995;Sweller et al., 2011b). If a concept is taught only through the visual channel such as text and picture or only through audio channel, it can potentially overwhelm a part of working memory without using the other parts to reduce the overload. Therefore, in order to increase the capacity of working memory and decrease cognitive load, both visually and auditory channels should be engaged simultaneously through presenting a part of the information in an auditory format and the rest in a visually format (Berney & Bétrancourt, 2016;Clark & Mayer, 2016;Tabbers et al., 2004).
Since E-learning systems provide a multimedia environment, it is important to strike a balance between visual and audio in order to teach software applications. To this end, we should try to use the software interface as the only visual element and minimize the amount of text as much as possible, and instead use audio to describe the tools in order to engage both auditory and visual channels and decrease cognitive load (Clark & Mayer, 2016;Tabbers et al., 2004). However, using long or complex audio content should be avoided as based on the transient information effect replacing long text with transient information such as audio can decrease learning performance, since verbal information may not be retained in working memory long enough to be fully comprehended. Segmenting longer or more complex audio content with pauses can help reduce transience related effects (Leahy & Sweller, 2011;Singh et al., 2012Singh et al., , 2017Wong et al., 2012).
It is the same story for animation, if animation is used for teaching lengthy content, it can overwhelm working memory load, since information from previous frames should be kept active in the working memory in order to understand later frames and enable learners to integrate the newly presented information with previously learnt information (Wong et al., 2012). Also, when animation is used for teaching concepts that learners need to spend more time on or focus on a specific part of the content such as learning words or symbols, the transient nature of the animation can increase cognitive load (Castro-Alonso et al., 2014).
In order to teach longer content using animation and audio, a segmentation strategy can be used to decrease the transient information effect (Kalyuga, 2008;Spanjers et al., 2011). However, it should be noted that the segmentation strategy is more effective for learners with low domainspecific knowledge. It means that segmentation will not be very effective for learners with more prior knowledge as they are able to deal with more information at any given time (Singh et al., 2012;Sweller et al., 2011a). Therefore, the transient information effect should be considered in using both animation and auditory information. For example, Figure 9(a) demonstrates a block of text that explains how to use formula in Excel. Since this text block is not long, it can be replaced by audio to use the auditory channel in addition to the visual channel (software interface) in order to decrease learners' cognitive load. Also, the formula section can be highlighted to provide signalling ( Figure  9(b)) and decrease the audio transient effect by keeping learners' concentration on the target teaching content (Alpizar et al., 2020).

Discussion
Based on what was discussed in this paper, in order to design an efficient E-learning system that can decrease learners' cognitive load there are various instructional design solutions that can be considered in different stages of teaching.
Based on the results of this review, teaching software requires breaking the teaching content into smallest chunks and ensuring these chunks are building on student's pre-requisite schemas for learning and where possible linking these chunks to software or features that are familiar for the learners. It is also paramount for the designers of teaching content to use strategies that create a mental connection between different features students have learnt, such as providing an infographic or detailed worked examples relative to the learning stage and relevant concepts, or that create connection to real-life situations, such as using a familiar narrative. These can support schema creation which in return, facilitates learning.
Using these guidelines alongside strategies that decrease learners' cognitive load leads to an optimum design that ensures learning. These cognitive load decreasing strategies include splitattention and redundancy effect avoidance, applying the training wheel method, and balanced visual and co-ordinated auditory information input. In what follows, we examine these strategies and provide specific examples for each strategy, and conclude by proposing a graphical summary of cognitive-based effects on learning software.
As the first step the learning method should facilitate the formation of schemas by providing necessary pre-required fundamental knowledge about the usage of the tools and by breaking down the instruction into the smallest teaching chunks.
Step-by-step instructions instead of a long description, especially for the tools with high interactivity, are suggested (Sweller et al., 2011a;Van Merriënboer, 1997). For example, in order to teach how to write a specific function in a spreadsheet application, first the usage of that function and its differences to similar functions should be taught, followed by step-by-step instructions of the parameters that should be added in each section of the function.
The next step to improve facilitating the formation of schemas is helping learners to connect new knowledge with their prior knowledge. As the first solution, we should try to link the layout and tools of the target software to the software that learners are already familiar with by using relevant analogies (Dahl et al., 2008;Gick & Holyoak, 1987;Quiroga et al., 2004;Sweller, 1994). For instance, Photoshop tools can be taught by comparing and connecting its tools to similar tools in Microsoft paint, and discussing the similarities and differences. Furthermore, applying worked example and integrating them using a familiar narrative is another efficient method that can facilitate the construction of schemas for the learners to enable them to solve relevant problems (Palomino et al., 2019;Pollock et al., 2002;Van Mierlo et al., 2012). Using real world examples and problems instead of just introducing the tools' usage can help learners to gain practical knowledge and extraneous cognitive load will be decreased especially when the target tool is difficult to learn (Atkinson et al., 2000;Clark & Mayer, 2016;J. J. Van Merriënboer & Kirschner, 2017;Sweller & Cooper, 1985). However, based on the expertise reversal effect, the effectiveness of worked examples can be decreased after learners become adequately familiar with the software. Therefore, we should eventually replace simple worked examples by more complicated examples that can demonstrate how to do a task using a combination of tools including partially or completely open-ended problem-solving exercises (Kalyuga & Renkl, 2010;Nievelstein et al., 2013;Renkl et al., 2004;Stark, 1998;Van Merriënboer et al., 2002). For example, in order to teach how the combination of different tools can be used, problem-solving exercises with minimal guidance can be used. This, of course needs to be after users have learnt all the related tools for polishing a face in Photoshop with the use of different practical worked examples. After the minimally guided problem-solving exercises stage, the minimal guidance can be replaced with the open-ended problem-solving exercises to fully support the development of expertise.
Lastly, providing an infographic at the end of each lesson can help to form schemas by showing a summary of the instructions in a step-by-step visual format (Clarke et al., 2006;Gobert & Clement, 1999;Lyra et al., 2016;Marcus et al., 1996;Yildirim, 2016).
In order to deliver the teaching content, one of the most effective solutions is using interactive animations to learn how to use the tools of graphical software applications. Since interacting with the software interface can increase learning performance by giving users practical activities during the learning process (Darejeh, 2021). Also, this interaction can help learners to easily remember the location of different tools within an interface, as in addition to watching an animation, they also click software tools which can increase learning performance based on embodied cognition effects, where mind body interactions can be used to support cognition (Shapiro, 2019;Weisberg et al., 2017;Wilson & Foglia, 2017).
In addition to the above solutions, another important factor that can decrease learners' cognitive load is keeping their concentration on the target tool that they are learning by avoiding-split attention and redundancy effects and applying the training wheel method. In order to decrease split-attention effect, the instruction steps should be integrated with the software interface and the steps placed next to the related part of the software interface (Al-Shehri & Gitsaki, 2010;Schroeder & Cenkci, 2018). The redundancy effect can be eliminated by avoiding extra nonessential graphics and decorative elements such as pictures, background music, or talking avatars (Bus et al., 2015;Craik, 2014;Jaeger & Wiley, 2014;Vossing et al., 2016;Wang et al., 2017). Lastly, the training wheel approach can be applied to only make visible the software tools that users have learnt in the previous lessons and the current tool that they are learning and to hide unfamiliar tools to ensure that they have complete focus on the target tool (Bannert, 2000;Fang et al., 2011;Pociask & Morrison, 2004;Reis et al., 2012). For example, in order to teach how to enter data in Microsoft Excel, all the tools in the font section can be deactivated and greyed out to keep learners' concentration on the worksheet and data entry process.
Finally, the E-learning systems should be designed in a way that there is a balance between visual and co-ordinated auditory information based on the modality effect, which emphasizes that there are separate processors for visually and auditory information. (Berney & Bétrancourt, 2016;Clark & Mayer, 2016;Tabbers et al., 2004). One should aim to reduce the amount of text and instead use audio to describe the software interface, with the aim to keep the audio short and simple to avoid the transient information effect (Leahy & Sweller, 2011;Singh et al., 2017;Wong et al., 2012).
Based on the different ways in which cognitive load can be affected while learning of software applications via E-learning systems, we proposed the framework below to increase the efficiency of E-learning systems for novice users (Figure 10). It should be noted that this framework would be only applicable when the learners are novices and they want to learn new software from scratch, however, for expert learners, this framework might not be beneficial.
We note that there might be other ways that cognitive load effects can be applied to the design of E-learning for teaching software applications, however this overview has focused on those that are considered to be the most relevant and practical to be implemented.

Conclusion
Cognitive load theory is based on extensive empirical evidence and significant number of studies have provided support for its effectiveness. Therefore, considering this theory can help us design more efficient E-learning systems which not only provide all the necessary teaching content but also facilitate learning using appropriate instructional design strategies. Designing E-learning systems for teaching software applications based on the principles of cognitive load theory, can support learning by adapting the learning content to our cognitive architecture, preventing working memory from being overwhelmed, facilitating the formation of schemas in long-term memory, and consequently increasing learning performance. In fact, this can lead to more efficient and reliable learning with less effort on the part of the learners. A summary of the solutions to decrease cognitive load of novice users in order for them to learn software applications through E-learning systems can be seen in Table 1.

Formation of Schema
Difficulty in learning high element interactivity tools. Avoid using decorative graphics when teaching software.

Modality and Transient information effects
Possibility of working memory being overwhelmed when teaching software using only visual or audio channel.
Engage both visual and audio channels by using the software interface as the only visual element and use short and simple audio to describe the tools.