Making Understanding Visible Exploring Mental Models with Concept Mapping

The ultimate goal of human-centered design is often simple to describe: Digital products and services that are intuitive because they simply work how they should work. However, what is simple to describe is often hard to achieve, and the key to a good user experience (UX) lies in UX research. UX researcher need to understand how users expect a digital product or service to work. Then, UX designers need to take these insights into account. This is where mental models, a concept from psychology, come into play.

How do mental models help with human-centered design?

Mental models represent our ideas about the environment, such as how a technology works. Our mental models are based on our experience. They can be incomplete or even outright wrong. We construct a mental model based on our understanding of the situation. When we learn something new or have an unfamiliar experience, we update our mental models (Nielsen Norman Group on mental models).

Mental models help to create positive user experiences with digital products and services. Users find it easy and straight-forward to interact with a product or service when their mental models matches how the interaction works. Otherwise, they have to learn how it works, which is cumbersome and frustrating. This could even reduce acceptance of a product or service. In summary, mental models can help create intuitive digital products and services. But how to consider them in UX design?

A good starting point is to observe conventions and established patterns of interaction design and information architecture, such as the typical position of a website's main navigation at the top or on the left. By doing so, users can readily apply their knowledge to a new product or service. However, this alone is often not enough: Mental models are much more than such conventions, and often the ideas of designers and users are different. After all, anyone who designs a digital product or service knows quite a bit about how it works. This makes it very difficult to imagine how the interaction feels to someone without this knowledge. For this reason, it is imperative to investigate mental models. The purpose of this article is to propose a method that is ideally suited for this investigation due to its structural nature and flexibility: Concept Mapping.

Basics of concept mapping

Concept mapping is a variant of so-called "cognitive maps", which aim at visualizing mental models. The method goes back to Joseph Novak and originated in education. In the 1970s, Novak's team wanted to help students to link new to existing knowledge and to better understand structural relationships. The method of concept mapping is one of the results of this research. Concept maps are structural visualizations of relationships and are based on three basic building blocks:

• Concepts represent persons, living beings, objects or events. They are usually written in shapes. Their names should be as clear and concise as possible.
• Links are connections between concepts and can be represented as lines or arrows. Most concept maps use labeled links to describe the connection as precisely as possible.
• Groups of concepts and links are called propositions. Most often, these are two concepts with a connecting link. Propositions between different branches of a concept map are particularly interesting because they show connections between separate areas. Such propositions are called cross-links or cross-references.

Novak's original idea has well-defined rules. Concept maps should be hierarchical and move from an abstract, general concept to concrete, specific concepts. Such a concept map has different branches that become more and more concrete from top to bottom. Propositions within a branch are represented with simple lines. Crosslinks between branches use arrows. Concept Maps in this original sense are well suited for static, structural knowledge.

Besides these hierarchical concept maps, there are also free, network-like concept maps. They represent associations without a defined structure. This makes them more flexible but also more complex, because the reading order and importance of the elements are not fixed.

Finally, there are variants specifically for dynamic topics. They use rounded arrows and specify how the linked concepts impact each other, for example with "+" and "-" signs. These concept maps are suitable for processes or complex problems. Let's look at an example: How could we increase awareness of phishing attempts in an organization? This is a complex question. We could simulate phishing attacks, and this could provide a higher awareness of phishing. However, simulated phishing attacks could also have a negative impact on employees' motivation.

Although this variety of concept maps could be confusing at first sight, it also makes the method of concept mapping very flexible. For example, hierarchical concepts maps are useful for structural topics, while a dynamic concept map is suitable for processes and complex problems. During user research, an experienced researcher could pick the appropriate variant and help users in representing their mental models in the preferable style.

Using concept maps in user research

The first key question to using concept mapping in user research is to decide how much freedom the participants have. In so-called directed concept mapping, you provide users with predefined content for their concept maps. This could be a list of terms for concepts and/or links, but also a skeleton concept map to enhance or a completed concept map with gaps to be filled. Such directed concept mapping makes it easier to compare concept maps from different people.

Directed concept mapping is useful for testing whether content helps users build better mental models. As an example, imagine you are working on a tool for encrypting email with PGP. PGP requires some technical understanding, and an encryption tool could work with an onboarding or tutorial to help users get started. In this case, you could use concept mapping to evaluate the success of the onboarding. Users first create a concept map on the focus question "How does email encryption work?" based on a few relevant terms (prior knowledge). After seeing the onboarding and interacting with the tool, they create another concept map with the same terms. Comparing the two concept maps allows you to see whether the users' mental models became more precise or whether there are still knowledge gaps. Afterwards, you can use these insights to optimise the design and content of the onboarding, and do another concept mapping with users as a quality check.

Less directed or undirected concept mapping provides users with few or even no guidelines to consider. This avoids influencing users and makes it easier to capture their individual mental models. On the backside, undirected concept maps also vary widely and are harder to compare. Such concept mapping is therefore suitable for complex topics with multiple layers. As an example, consider the topic of how much data we should collect at the workplace. We could use sensors to create profiles of the current situation, for example about the air quality. Would this be appropriate? Would employees accept such sensors? Would it depend on the kind of sensor and which specific data we are collecting? This is a typical example of a complex topic with multiple layers: We could use the collected data to monitor employees or to improve their well-being (for example, by encouraging ventilation). An undirected concept mapping on the focus question "Should sensors be allowed to collect data about the situation at the workplace?" allows employees to express their views about what to consider in the decision. This also works well in a semi-structured interview, when a UX researcher creates a concept map together with the participants. You could then always check whether the concept map correctly represents the user's mental model and ask for further explanations if necessary.

Concept mapping is relatively easy to learn. However, here are some tips and tricks for practically applying concept mapping in user research (Novak & Gowin, 1984; Novak, 2010):

1. Clearly define the topic: First, define the topic of the concept map as clearly as possible. It is preferable to use a so-called focus question that explicitly defines what the concept map wants to answer. The wording of the focus question has a great impact on the content of the concept map. A focus question starting with "What is..." will stimulate factual knowledge, while a focus question with "How does..." will tend to stimulate process knowledge. Depending on your interest, you can provide the participants with the focus question or ask them to define them themselves.
2. Collect terms: The next step is to collect 10-20 relevant terms and arrange them in a meaningful structure on the concept map. You should place related terms close to each other.
3. Create links: Next, create links between the terms and write labels on the links. This creates the "propositions" mentioned above. Propositions should be clear and concise. They should also be independent of each other because the reading direction is not fixed in concept maps.
4. Perform a quality check: Finally, you should review the concept map and check its quality. Does it represent the topic adequately, or is there content to add or change? Often, the elements in a concept map are extensively changed or restructured in this phase.

Of course, this process is idealized, and there are often deviations in practice. Participants often rethink or reorganize content during concept mapping. This active reflection is a great advantage of concept mapping because it helps to become aware of the structure of one's mental model.

Sometimes, it is useful to create concept maps repeatedly over a longer period of time. In this case, the concept maps visualise how mental models develop over time. For this purpose, UX researchers organize regular sessions with participants. In each session, they modify concept maps from previous sessions. In a concept mapping study, the researchers Camille Kandiko and Ian Kinchin used such an approach to investigate how the mental models of doctoral students change over time while working on a doctoral thesis. While the concept maps focused strongly on the "product" of the doctoral thesis at the beginning, later concept maps showed a stronger focus on the process of doctoral studies itself.

Better UX design through concept mapping

You can analyse concept maps according to different criteria, such as the accuracy of the content, structural aspects, or even network analysis. In UX research, it is a good idea to focus on propositions and structures. Propositions are the meaningful units in concept maps and define relationships. Are there patterns in the concept maps that indicate ambiguity (for example, because propositions are inaccurate or unlabeled)? Are there isolated areas that indicate that users are unable to see the connections between different areas?

Another useful method is to look at the structure of a concept map. There are several typical structures (Yin et al., 2005):

• Chain structures represent linear processes with a clearly defined start and end.
• Circle structures represent feedback loops. They are typical for topics where one concept reinforces another.
• Tree structures are hierarchically organized, similar to a taxonomy.
• Star- or spider-shaped structures have a clearly defined center which acts as a bridge to all related areas. They show some understanding of interrelationships between different concepts, but are not yet differentiated.
• Finally, network structures represent highly integrated mental models with strong interconnections between different areas.

Note that there is not one universally "correct" structure: You always have to relate structure of a concept map to the specific question it is trying to answer. For example, a chain structure may be just right for a step-by-step process. However, if the concept map represents a complex topic, a chain structure may stand for a mental model that is far too simple.

After analysing the concept maps, you can draw relevant conclusions for UX design. First, the concept maps of different users can be compared with each other. Are they similar or very different? Is there information that could explain these differences, for example, different previous experiences with technology? For example, different mental models might indicate that there are different user groups who need different design solutions.

You could also ask the designers to create concept maps about how the product or service works. Comparing the designers' concept maps to the users' concept maps can help to analyse whether the mental models match. For example, if the intended functionality does not match the mental models of the users, a more intuitive design is needed in the next iteration. If that is not possible, a design might need assistance to help the users adapt their mental models. And a new round of tests, maybe again via concept mapping, helps to monitor success.

Conclusion

Mental models are important in UX design: They help to create intuitive digital products and services that respond appropriately to users' needs. Concept maps are a flexible way to visually capture these mental models, communicate about them, and identify changes. With this knowledge, UX designers can modify digital products and services or help users build mental models to improve the user experience.

---

Note: A version of this article in German was published in t3n 68.