Chapter 8. Technologies for spatial learning

Kenneth D. Forbus
David Uttal
Northwestern University

Paper-based technologies, such as maps, diagrams and graphs, have been used to support spatial thinking and learning through the ages. Modern digital technology provides new options for enriching spatial learning with potentially transformative impacts. This chapter focuses on two such technologies. The first is sketch understanding, where software incorporates cognitive models of human visual, spatial and conceptual representations and processes in order to see a student’s sketch as they do. We summarise CogSketch, a new platform that has been used to develop two novel kinds of educational software that have been successfully used in classroom experiments. The second technology are Geographic Information Systems (GIS), computer-based mapping systems that facilitate spatial data analysis and visualisation, e.g. the distribution of housing density, crime, school test scores, etc. in different neighbourhoods. We review evidence that using GIS can promote spatial reasoning in high school students.


Context statement

Spatial learning – learning to reason better about space and using space to learn about other subjects is critically important for science, technology, engineering and mathematics (STEM) education and practice. However, spatial learning has not been emphasised in education, in part because of the challenges of creating and assessing multiple representations of space. We show that new technologies can overcome these challenges. Modern digital technology provides new options for enriching spatial learning (Jee et al., 2014[1]). Here we focus on two technologies. The first is sketch understanding, where software incorporates cognitive models of human visual, spatial and conceptual representations and processes in order to see a student’s sketch as they do, and thereby provide immediate feedback. We summarise research on CogSketch, a new platform for sketch-based educational software that has been used to develop two kinds of educational software that have been used in successful classroom experiments. The second is the use of Geographic Information Systems (GIS), computer-based mapping systems that allow the user to examine data in layers. We review evidence that using GIS can promote spatial reasoning in high school students.

Spatial technologies

People have been using external media to support spatial learning for millennia, going back to the construction of sketches in dirt and making models out of found materials. Today’s digital technologies provide several key advantages for spatial learning:

  1. 1. Access to vast reference sources. The web, combined with open-source and open-data movements, has made a cornucopia of information widely and easily available for those with internet connectivity. The semantic web is adding a layer of metadata that further supports finding and using resources in novel ways.

  2. 2. Simplified sharing. Digital versions of maps, diagrams and sketches can be easily copied and distributed online, compared to copying and distribution of paper products. This facilitates students sharing work with each other and producing more complex materials for assignments. For example, digital sketches can include a complete history of a student’s actions, which can be used for further assessment, which is impossible with paper sketches.

  3. 3. Intelligent software coaches and tutors. Rapid feedback is known to facilitate learning. With pencil and paper, feedback is delayed until a human instructor can look at it, and send it back to the student, often days, weeks or even months later. Digital media can have embedded software coaches and tutors, capable of providing helpful feedback immediately, anytime, anywhere (see also Chapter 12, by Klahr and Siler, and Chapter 13, by Koedinger).

There are many technologies that could have important impacts on spatial learning, including virtual reality, augmented reality and gesture understanding. Here we focus on two technologies that are now translation ready and have shown potential for improving spatial learning: Sketch understanding and GIS. We discuss each in turn.

Sketch understanding

Teachers and students sketch to communicate ideas to each other, and to think through problems. Creating software that can understand sketches in ways that people do is the goal of sketch understanding research. Recent progress in sketch understanding has enabled the construction of new kinds of intelligent software to support spatial learning.

What is sketch understanding?

There is a common misconception that sketch understanding can be equated with sketch recognition. Sketch recognition has been used in domains where there is a vocabulary of abstract visual symbols that are part of domain practice, such as electronics (de Silva et al., 2007[2]), chemistry (Pargas et al., 2007[3]) and structural mechanics (Valentine et al., 2012[4]). However, there are two reasons why sketch understanding is more complicated than that. An analogy with speech recognition, which more people have experienced today, is helpful. Just because a system can transcribe spoken words into text does not mean it understands what those words convey (any user of Siri, Alexa or Cortana has experienced this). By contrast, people can solve complex visual problems, requiring rich representations and processing. Sketch understanding must investigate these visual and spatial representations and processes. The second reason is that most things that people sketch are not visual symbols. A speech recognition system produces garbage when fed music or street sounds – most of the sounds in the world are not human speech. And so it is with sketching: the spatial aspects of most things depicted in STEM domains are determined by the nature of the situation, not by symbolic conventions. The mapping between shapes and concepts is many to many, even within single domains. For example, circles in an Earth Science course can be used to depict the layers of the earth, planets and some orbits. Hence the goal of sketch understanding must include ascertaining what conceptual relationships follow from the visual and spatial relationships depicted within a sketch.


This insight that sketch understanding requires integrating conceptual, visual and spatial representations and reasoning has led to CogSketch (Forbus et al., 2011[5]), a software system that is both a computational model of aspects of human visual understanding and a platform for sketch-based educational software. Motivated by studies of human vision, CogSketch uses multiple, hierarchical levels of visual representations. Its initial descriptions are in terms of visual objects (called glyphs). Figure 8.1, for example, there are three glyphs, called Block, Ramp and Gravity. CogSketch can decompose glyphs into visual edges (e.g. four for the Block) and uses gestalt principles to flexibly break apart and combine visual materials. These capabilities have enabled CogSketch to model multiple visual tasks, including geometric analogies (Lovett et al., 2009[6]), an oddity task (Lovett and Forbus, 2011[7]), paper folding and mental rotation. Evidence of the fidelity of CogSketch’s representations and reasoning can be seen in the ability of these models to provide explanations for human performance, including predictions of ordinal reaction times and variations in problem difficulty. For example, the performance of the CogSketch model of Ravens’ Progressive Matrices (Lovett and Forbus, 2017[8]) places it in the 75th percentile, making it better than most adult Americans. These visual representations and reasoning capabilities are used in the educational software discussed next. In addition to visual representations, CogSketch uses the contents of Cycorp’s OpenCyc knowledge base, which includes over 58 000 concepts, 8 000 relationships, whose meanings are specified by roughly 2.3 million facts. This background knowledge is augmented by qualitative representations for mechanics and other substrate capabilities needed to support STEM domains, although as discussed below, these capabilities will need to be built out considerably to handle STEM education more broadly.

Figure 8.1. A simple Design Coach example
Figure 8.1. A simple Design Coach example

Note: Given an explanation that does not fit with its conceptual understanding of mechanics, the system provides feedback about what parts of the explanation do not make sense.

To see how these kinds of knowledge interact, let us examine an example of an educational software system built on CogSketch, the Design Coach (Wetzel and Forbus, 2009[9]). The Design Coach attempts to address a problem that instructors in the Design Thinking and Communications class at Northwestern have with their students. While the instructors view sketching as a crucial way to think through new designs and communicate them to clients, they find that students are often afraid to sketch, and become embarrassed about their drawings. Design Coach provides a safe way for students to practice explaining designs through sketching. They use a combination of drawings and language-like input to explain their ideas. The Design Coach uses spatial reasoning and qualitative mechanics to look for problems or gaps in their explanations, and gives them feedback when it does not understand their design. Students can then change their explanation until the system understands it. Pilot data indicates that even doing a single Design Coach assignment can reduce student anxiety about communicating via sketching (Wetzel and Forbus, 2015[10]). Let us walk through a simple example. Suppose part of a student’s design has a block sliding on a ramp (Figure 8.1). They draw the ramp, and give it a conceptual label of Fixed Object, indicating that this is something that cannot move. The conceptual label is applied by a simple menu interface, thereby sidestepping all issues with sketch recognition. The glyph for the block is labelled as a Rigid Object (by contrast with a cord or spring). The visual overlap between the two glyphs means, given their conceptual interpretations, that the two objects have a surface contact. Visual processing on the student’s ink automatically identifies this region of the sketch. This visual analysis, combined with its conceptual knowledge, enables the Design Coach to reason through, in a human-like way, what might happen next. Using the qualitative orientation of the surface and the direction of gravity, as depicted by the arrow the student drew, CogSketch determines that the block will slide to the left and down. If this is what the students said in their explanation, the Coach tells them that it understands their explanation. If instead they said that the block would move in a different direction, it would point out the difference between what they said and its understanding of what might happen, as Figure 8.1 illustrates. Using qualitative, conceptual reasoning to check student reasoning instead of, say, numerical simulation, is important because it enables the Design Coach to explain its reasoning to students.

Sketch Worksheets

The Design Coach was targeted for a specific problem within a particular area. The second kind of educational software built on top of CogSketch are Sketch Worksheets (Forbus et al., 2017[11]; Yin et al., 2010[12]). Sketch Worksheets are domain-general. Their goal is to help students learn spatial layouts or domain information that can be communicated via visual representations, such as concept maps. To achieve generality, tutoring in Sketch Worksheets only relies on analogical matching between an instructor’s solution and a student’s solution. Each Sketch Worksheet is focused on a specific problem. The instructors (or curriculum designer), who sets the problem uses, CogSketch to draw their solution to that problem. They select which concepts and relationships from the knowledge base will be available to the student doing that exercise, and control how those concepts are displayed to the student. This enables the worksheet author to keep the student focused on the key concepts (plus distractors, typically) relevant to the exercise at hand, and tailor the description of the contents to be age and topic appropriate, as well as in the student’s native language. CogSketch’s visual system automatically analyses their solution, and authors browse through natural language depictions of the facts it derived. They can select some of these facts as important, i.e. if they are not true of a student’s sketch, then there is something wrong with it. They can supply feedback to give to students in that case, as well as provide rubrics specifying how many points each important fact is worth. When students tackle a worksheet, they draw their sketch, using the set of concepts and relationships specified by the teacher to label their ink. When asked for feedback, CogSketch analyses the student sketch, and compares this analysis with its analysis of the solution sketch (to do this comparison, it uses a computational model of analogical comparison, the same model of analogy used in the visual reasoning models outlined earlier). The text associated with any differences found are provided to the student for feedback, highlighting the involved glyphs to help students see how the feedback applies to their sketch. If there is no feedback, the worksheet tutor congratulates the students on their sketch and they are finished. Students sometimes do stop early, but the ability to provide rapid feedback anywhere at any time is one of the valuable aspects of this model. Instructors can also provide misconception sketches that highlight common problems with student mental models, and give feedback based on the comparison with those sketches instead, when they match. CogSketch maintains a complete history of student actions, including what the sketch looked like every time a student asked for feedback, which can be used for additional assessment and data-mining (Chang and Forbus, 2014[13]).

Sketch Worksheets have been used in a variety of laboratory and classroom experiments. For example, students in a fifth-grade biology class showed gains in two out of three pre-test/post-test pairs after doing three worksheets on the human circulatory system (Miller, Cromley and Newcombe, 2013[14]). A set of worksheets for introductory geoscience courses were developed by researchers at the Spatial Intelligence and Learning Center of the University of Wisconsin, Madison and used in classes there and at Northwestern University. While geoscience instructors think sketching is important, they do not give frequent sketching assignments due to the burden of grading paper-based sketches. The Madison experiment did not find differences in learning between paper-based and CogSketch-based worksheets, while CogSketch-based worksheets are far easier to grade (Garnier et al., 2017[15]). The geoscientist-authored worksheets from Madison were subsequently deployed in Northwestern classes, and are now available on line from the NSF-funded Software Engineering Research Center (SERC). Moreover, Sketch Worksheets have also been used in other Northwestern classes, in Knowledge Representation and Introduction to Cognitive Modeling (Forbus et al., 2018[16]). These experiments and deployments suggest that Sketch Worksheets could potentially be valuable across a wide range of STEM topics and age ranges.

Figure 8.2. An example of a Sketch Worksheet about the greenhouse effect, with feedback
Figure 8.2. An example of a Sketch Worksheet about the greenhouse effect, with feedback

Geographic information systems (GIS)

GIS technology has transformed the presentation of spatial information in many fields, ranging from marketing to architecture. A GIS is a database of spatial information, where entities in the world, or conceptual entities computed about the world (e.g. population density) are represented by polygons. These polygons are organised into layers, much like acetate overlays are used with maps. Defining and manipulating layers provides a valuable tool for spatial thinking. For example, a user can represent the locations of parks, businesses and schools within a particular city. Each layer can be added or subtracted as desired, allowing the user to see, and to think about, relations both within individual layers and among different layers. One use of GIS is to ascertain good locations for specific purposes, e.g. where to put a new fire station. The invention and popularisation of GIS led to substantial growth in the availability and sharing of spatially-coded data. For example, many US cities now make available GIS data concerning zoning, school catchment areas, and crime distributions.

There are many forms of GIS. Professional GIS, such as ArcGIS for Desktop (ESRI, 2014) are extremely powerful, providing numerous tools for representation, animation, and transformation of images. Such systems can be hard for novices to learn. Many simpler tools now provide some of the same services in a more approachable form, such as Google Maps and Google Earth.

GIS in education

A GIS supports students defining and solving spatial problems by gathering relevant data, decomposing it into layers, and manipulating them. This makes it especially effective for emphasising relations, patterns, and distributions (Sinton et al., 2013[17]). This approach to teaching and learning is consistent with recent approaches to science education that emphasises science as a set of practices. These practices include collecting data, representing and forming models, iterating, etc. (NGSS Lead States, 2013[18]). We suggest that GIS facilitates many of these processes, as it encourages thinking about data, representing and modelling, and combining representations to solve spatial problems.

Here we highlight two approaches that we believe illustrate 1) the unique affordances and instructional advantages of using a GIS-based approach to teach spatial problems; and 2) two different models for GIS use. One approach uses free tools (e.g. Google Earth) to teach a particular unit within Earth Science, while the other (the Geospatial Semester) uses a professional GIS to emphasise a problem-solving approach. We discuss each in turn.

The first approach uses GIS within specific topics or units within courses, such as Earth Science. A good example of this approach is (Bodzin, 2008[19]; Bodzin and Cirucci, 2009[20]), use of Google Earth to teach a variety of environmental science topics. For example, in one study (Bodzin and Cirucci, 2009[20]), students used Google Earth to investigate land use and how changes in land use affect the local environment. For example, students investigated how the building of a local shopping mall affected the pattern of land use around the mall. They then investigated how, and why, the presence of the mall led to heat islands, which are concentrated areas of relatively high temperature. The loss of plant cover, and its replacement with concrete, asphalt, etc., led to greater accumulation of heat and slower loss of heat after dark. This helped students learn how to think about complex, real-world spatial problems.

The second approach uses a more intensive approach to learning through GIS usage, working on more open-ended problems. For example, the Geospatial Semester (GSS) (Jant, Uttal and Kolvoord, under review[21]; Kolvoord et al., 2012[22]), is a semester or year-long class designed primarily for high school seniors. It emphasises a spatially-based approach to real-world problem solving. In contrast to the more specific curricula discussed above, the GSS emphasises a more holistic approach to learning and problem solving. Students learn to use the inexpensive educational version of ArcGIS for Desktop, a full GIS that allows flexible representations of spatially-coded data. The first portion of the course is devoted to learning the mechanics of ArcGIS and using it to solve specific problems. But as the course progresses, students begin to think about new problems that can be solved with GIS. The course culminates with an intensive, multi-week final project. Students must identify a significant problem that can be solved using GIS technologies. The students work to identify the problem, constrain possible solutions, identify relevant data, and propose a solution. The problems often involve engineering, in that they are forced to make compromises and offer working solutions rather than a perfect solution.

Some examples of final projects provide illustrations of the approach. One project, Bearly Relocated, was completed in a rural high school located near Shenandoah National Park in Northeast Virginia. The problem begins with the geography of the park itself; it is relatively easy for bears to leave the park in search of food. The bears have learned that it is easy to obtain food by venturing out of the park (where they are safe and protected) and enter a residential or farm area (where they put both themselves and the local human population at risk). It is not uncommon to find a bear in one’s backyard. Thus, students were motivated to find a way to solve the problem. Working with a GIS allowed the students to come up with possible solutions. The basic idea was to identify areas of the park in which the bears could be relocated to decrease the chances of them leaving the safety of the park again. This requires evaluating trade-offs among a variety of variables: The locations where the bears are found out of the park, the presence of access roads, etc. The students use ArcGIS to define optimal areas within the park and shared these recommendations with the park rangers.

Another example of a compelling final project was investigating how to increase internet access in Africa. Lack of infrastructure and vast distances between large cities makes the problem particularly severe. Radical new solutions, such as balloons or drones, have been proposed to provide wireless access. The students tackled the question of where such systems could be placed to provide maximum benefits with minimal costs. This is a highly spatial engineering and economic problem, involving many variables. For example, some regions are already well-covered (e.g. urban areas in South Africa), while others are so desolate that the number of people served by covering them would be small. Political stability is a factor, e.g. a country engaged in a civil war is unlikely to maintain the system. To generate possible solutions, the students explored these factors and others using a GIS. They constructed representations of population density, political stability, and the physical requirements of the airborne system, to ascertain which countries would be the highest-pay-off investments, and where airborne systems should be positioned.

Systematic research on the effectiveness of GIS

Examples such as these highlight the potential for interventions like the Geospatial Semester that use a GIS intensively to support student spatial learning and problem solving. However, additional research is needed to investigate whether these effects hold up at a larger scale. Do students who enrol in the GSS learn to think more spatially and more effectively than students enrolled in other classes?

To address this question, we (Jant, Uttal and Kolvoord, under review[21]) conducted a quasi-experiment, comparing students’ performance in the GSS to that of students in two other demanding senior-level courses, AP Physics and AP History. We interviewed students several times across the year, asking about their developing ideas for the final project, and what they had learned. In addition, we also asked the students a series of transfer questions, in which students were asked to solve new problems. The goal of the transfer questions was to assess whether students considered spatial approaches to new problems. For example, one transfer problem asked students to imagine that they were running for sheriff in their local county, and to think about how they would go about soliciting votes. This problem is not inherently spatial; it could be solved simply by saying they would run advertisements in local newspapers or through direct mail. But an effective solution will often involve thinking about different layers of spatial data, including population density, outcomes of prior elections in given areas, polling locations, etc. Thinking of the problem in a spatial manner allows one to be much more selective and efficient about where and when to allocate effort and money in the campaign.

We found that students in the GSS class approached the transfer problems in a spatial manner, thinking about the problems as involving distributions and patterns. They mentioned the kinds of data that they would need to locate and represent to solve the problem. They also recognised the iterative nature of problem solving, noting that the GIS would allow them to consider information in multiple ways. The results suggest that working with GIS can substantially alter both how students conceive of and implement possible solutions.


Digital versions of paper technologies offer new opportunities to support and enhance spatial learning. For example, maps have been used for thousands of years, and learning to use them is an important aspect of the development of spatial cognition (Uttal, 2000[23]; Uttal and Sheehan, 2014[24]). Maps help us to think about the world beyond direct experience and to learn about the world from others. GIS expands this basic capacity of maps by allowing the user to select and represent a large variety of different data, in different layers. As we have discussed, GIS enhances spatial thinking by allowing the user to combine information in ways that would be very difficult if not impossible to do on paper. With training and experience, GIS has the potential to expand students’ conceptions of problems and their possible solutions; learning to use GIS helps students to conceive of spatial solutions to problems, and to explore a range of possible solutions to these problems.

Similarly, software that understands sketches in human-like ways make it possible to create software coaches and tutors that can provide students with immediate feedback anywhere, anytime. While CogSketch already has enough visual representations and reasoning to support new kinds of educational software, there is much research ahead before the goal of human-capable visual understanding is reached. Significant progress in at least four areas should be goals of future research:

  1. 1. Richer 3D understanding. The quantitative aspects of 3D modelling are well understood from prior work on computational geometry and graphics. But capturing the qualitative aspects of how people reason through three dimensional problems is still an open issue (Gagnier and Shipley, 2016[25]).

  2. 2. Deeper linkages to qualitative mental models. Spatial representations are typically carriers of conceptual information, and our understanding how visual properties of sketches are used to depict and reason through such conceptual information is only fragmentary at this point (Chang, Wetzel and Forbus, 2014[26]). Expanding the use of sketch understanding across multiple STEM domains would be a productive way to drive such research.

  3. 3. Integration with other modalities. People talk when they sketch with each other, so the fluency of sketching would be increased by supporting natural language, speech, and gesture in conjunction with the drawing component of sketching. Integrating sketch understanding with low-level visual analysis of images would enable systems to gain background knowledge from diagrams on the web, and work with student drawings produced outside of sketching software.

  4. 4. Multimodal memory systems. Creating intelligent software tutors and coaches that recognise common patterns of misconceptions, track students’ progress by comparing their work on novel assignments, and providing useful examples and precedents for projects that a student is working on, all require systems with libraries of experiences that consist of integrated visual, spatial, and conceptual knowledge.

Policy implications

Our research has several international policy implications. These could be enacted almost immediately, and many would be very inexpensive.

  1. 1. Spatial learning is critically important for STEM education. We now know that spatial learning is a very important part of STEM education and practice.

  2. 2. Spatial thinking is malleable; it responds to training and life experiences. We must design education to teach spatial thinking explicitly to see benefits in terms of STEM participation and achievement.

  3. 3. Spatial thinking and learning can be fostered greatly by new technologies, such as CogSketch and GIS. The technologies that we have reviewed here allow teachers and learners to think spatially in powerful, new ways.

  4. 4. These technologies are either free or inexpensive; cost should not be a reason to avoid using them. CogSketch (and associated worksheets) is freely available; all that is required to use it today is a Windows computer, and within the coming year a cloud-based version that can be accessed from tablets and smartphones will become available.

  5. 5. Sophisticated, cloud-based GIS programmes are also now available free of charge for schools. Thus, the time is ripe to begin using these spatial technologies more intensively in schools.


[19] Bodzin, A. (2008), “Integrating instructional technologies in a local watershed investigation with urban elementary learners”, The Journal of Environmental Education, Vol. 39/2, pp. 47-58,

[20] Bodzin, A. and L. Cirucci (2009), “Integrating geospatial technologies to examine urban land use change: A design partnership”, Journal of Geography, Vol. 108/4-5, pp. 186-197,

[13] Chang, M. and K. Forbus (2014), “Using analogy to cluster hand-drawn sketches for sketch-based educational software”, AI Magazine, Vol. 35/1, p. 76,

[26] Chang, M., J. Wetzel and K. Forbus (2014), “Spatial reasoning in comparative analyses of physics diagrams”, in Freksa, C. et al. (eds.), Spatial Cognition IX, Springer International Publishing, Cham,

[2] de Silva, R. et al. (2007), Kirchhoff’s Pen: A Pen Based Circuit Analysis Tutor, ACM Press, New York,

[11] Forbus, K. et al. (2017), “The cognitive science of sketch worksheets”, Topics in Cognitive Science, Vol. 9/4, pp. 921-942,

[16] Forbus, K. et al. (2018), Sketch Worksheets in STEM Classrooms: Two Deployments,

[5] Forbus, K. et al. (2011), “CogSketch: Sketch understanding for cognitive science research and for education”, Topics in Cognitive Science, Vol. 3/4, pp. 648-666,

[25] Gagnier, K. and T. Shipley (2016), “Visual completion from 2D cross-sections: Implications for visual theory and STEM education and practice”, Cognitive Research: Principles and Implications, Vol. 1,

[15] Garnier, B. et al. (2017), “Promoting sketching in introductory geoscience courses: CogSketch geoscience worksheets”, Topics in Cognitive Science, Vol. 9/4, pp. 943-969,

[21] Jant, E., D. Uttal and R. Kolvoord (under review), “Spatially enriched curriculum improves students’ critical thinking and spatial reasoning”, in Spatial Cognition 2014: Poster Presentations, 52.

[1] Jee, B. et al. (2014), “Drawing on experience: How domain knowledge is reflected in sketches of scientific structures and processes”, Research in Science Education, Vol. 44/6, pp. 859-883,

[22] Kolvoord, R. et al. (2012), “The impact of geospatial technologies in a dual-enrollment environment”, Society for Information Technology & Teacher Education International Conference, Vol. 1, pp. 1915-1922.

[8] Lovett, A. and K. Forbus (2017), “Modeling visual problem solving as analogical reasoning”, Psychological Review, Vol. 124/1, pp. 60-90,

[7] Lovett, A. and K. Forbus (2011), “Cultural commonalities and differences in spatial problem-solving: A computational analysis”, Cognition, Vol. 121/2, pp. 281-287,

[6] Lovett, A. et al. (2009), “Solving geometric analogy problems through two-stage analogical mapping”, Cognitive Science, Vol. 33/7, pp. 1192-1231,

[14] Miller, B., J. Cromley and N. Newcombe (2013), “Using CogSketch to support student science learning through sketching with automatic feedback. AERA NGSS lead states”, in Next Generation Science Standards: For States, By States, The National Academies Press, Washington, DC.

[18] NGSS Lead States (2013), Next Generation Science Standards: For States, By States, The National Academic Press, Washington, DC.

[3] Pargas, R. et al. (2007), OrganicPad: A Tablet PC Based Interactivity Tool for Organic Chemistry, IEEE,

[17] Sinton, D. et al. (2013), The People’s Guide to Spatial Thinking, National Council for Geographic Education, Washington, DC,

[23] Uttal, D. (2000), “Seeing the big picture: Map use and the development of spatial cognition”, Developmental Science, Vol. 3/3, pp. 247-286.

[24] Uttal, D. and K. Sheehan (2014), “The development of children’s understanding of maps and models: A prospective cognition perspective”, Journal of Cognitive Education and Psychology, Vol. 13/2,

[4] Valentine, S. et al. (2012), “Mechanix: A sketch-based tutoring system for statics courses”, in Proceedings of the 24th Conference on Innovative Applications of Artificial Intelligence, Texas A&M University, College Station, TX,

[10] Wetzel, J. and K. Forbus (2015), “Increasing student confidence in engineering sketching via a software coach”, in T. Hammond et al. (eds.), The Impact of Pen and Touch Technology on Education, Springer, New York,

[9] Wetzel, J. and K. Forbus (2009), “Automated critique of sketched designs in engineering”, in Proceedings of the 23rd International Workshop on Qualitative Reasoning,

[12] Yin, P. et al. (2010), “Sketch worksheets: A sketch-based educational software system”, in Proceedings of the 22nd Annual Conference on Innovative Applications of Artificial Intelligence,

End of the section – Back to iLibrary publication page