How Much Thinking Can a Team Handle?

Dieser Artikel ist auch auf Deutsch verfügbar

This article is part of a series

Part 1: Common Approaches in the Field of Socio-Technical Architectures
Part 2: Platforms, Teams, and APIs: How Do They Fit Together?
Part 3: Socio-Technical Architecture as a Competitive Advantage
Part 4: Don’t Forget the People
Part 5: How Much Thinking Can a Team Handle? (this article)
Part 6: Internal Development Platforms
Part 7: Enabling Stakeholders as a Success Factor
Part 8: Socio-Technical Architectures: Informality from Mining to Today[1]

The Role of Cognitive Load in Sociotechnical Systems

It’s reasonable that Cognitive Load is relevant in IT, particularly in software engineering. Software engineering is knowledge work performed primarily through thinking and communication—cognitive processes. Starting with the premise that high Cognitive Load is problematic, we can ask: How can Cognitive Load be reduced in an IT organization?

The Cognitive Load Theory

The concept of Cognitive Load, formulated by John Sweller and colleagues, addresses how much mental effort a person must expend to learn something. Sweller distinguishes three types of Cognitive Load that are fundamental to modern educational theory and can be applied to software engineering.

Intrinsic Cognitive Load represents the inherent complexity of a task: Adding two numbers is inherently less complex than performing polynomial interpolation. Consequently, the Intrinsic Cognitive Load for learning addition is lower than for learning polynomial interpolation. The Intrinsic Cognitive Load of a subject matter is inherent and cannot be altered.

However, it’s possible to build up the Intrinsic Cognitive Load gradually through different presentation approaches. For example, elementary school children first learn to add numbers from 1 to 10, then from 1 to 100. After mastering these concepts, the entire number range is introduced, and addition is applied as a general principle.

In essence: The more aspects relevant to a subject, the higher the Intrinsic Cognitive Load.

This occurs because successful learning requires simultaneously holding these aspects in working memory to connect them. By appropriately structuring learning material, abstraction levels for coherent sub-aspects can be introduced, allowing learners to progress incrementally—but this doesn’t reduce the actual Intrinsic Cognitive Load of the subject.

Extraneous Cognitive Load describes additional burdens created outside the learning subject. This stems not from the subject itself but from how the material is prepared and the environment in which learning takes place. This is particularly important for multimedia learning materials, as both movement and sounds consume working memory capacity and can therefore undermine the presentation described above. Examples include videos where the presenter is visible despite their visibility not contributing to understanding the content, or footnotes that contain relevant information but interrupt the reading flow.

Extraneous Cognitive Load can be managed relatively well, although reducing it often requires effort. A quiet environment, appropriate and consistent visualization, and a clear thread in the didactic structure are measures that can influence Extraneous Cognitive Load.

Finally, Germane Cognitive Load describes the effort of connecting elements in working memory with previously learned elements in long-term memory. Germane Cognitive Load essentially represents the “aha moment” when comprehension occurs. According to Sweller and colleagues, this comprehension becomes easier when Intrinsic and Extraneous Cognitive Load are lower. Learners can primarily only experience Germane Cognitive Load. Differentiating between Intrinsic and Extraneous Cognitive Load and working with these concepts is only possible for those who prepare learning materials.

Cognitive Load in Software Engineering

In software engineering, learning occurs with four specific subject areas: First, when learning a business domain; second, when understanding what a specific requirement means; third, when learning tools (programming languages, frameworks, etc.); and fourth, when comprehending unfamiliar code.

Each of these four learning areas carries an Intrinsic Cognitive Load according to the model presented above.

Business domains have their own complexity, which varies considerably. Depending on how many aspects of such a domain a software engineer has already mastered, learning a new one becomes easier.

Consequently, switching between similar business domains is more manageable: The remaining Intrinsic Cognitive Load when switching from balance sheet accounting to payroll accounting is lower than when switching to smart home control. The Intrinsic Cognitive Load of the three domains doesn’t change, but the Germane Cognitive Load for learners differs significantly.

New requirements also have an Intrinsic Cognitive Load and must be understood before implementation. Unlike learning the business domain, new requirements appear more frequently and in smaller pieces. Understanding the business domain can be considered a prerequisite for understanding the requirements:

If both the business domain and the requirement are new, the Germane Cognitive Load of both learning subjects increases, as they each function as Extraneous Cognitive Load for the other.

Learning new tools is a frequent necessity in software engineering. The impact on the Intrinsic Cognitive Load relevant to learners is influenced by familiarity with similar tools. Thus, the remaining Intrinsic Cognitive Load when learning a new version of a familiar framework is generally lower than when learning a completely unfamiliar framework that essentially performs the same task.

On the other hand, certain tools cannot be learned in isolation: It’s nearly impossible to learn a concept like Object-Oriented Programming without also learning a programming language.

The challenge in presenting these two learning subjects is to structure the learning path so that the Intrinsic Cognitive Load of both can be managed without requiring learners to make too strong a connection between them.

While a basic connection is necessary, too strong a connection would be undesirable because it would mean learners could only apply both subjects together and would have to completely relearn Object-Oriented Programming with a different language.

Finally, Cognitive Load can also be observed when understanding unfamiliar code. If the programming language and the requirement implemented in the code are known, the remaining Intrinsic Cognitive Load is lower than if one or both are unknown.

Additionally, the way the code is structured creates an Extraneous Cognitive Load that affects comprehensibility.

While formatting can be automatically adjusted, elements such as naming and comments can either increase or reduce.

The same applies to the use of design patterns.

Cognitive Load in Teams

The Cognitive Load Theory has been developed over four decades and has been repeatedly validated for individual learning.

With the increasingly complex tasks related to digitalization and interdisciplinary problems, the question arises to what extent it is also valid and applicable for teams. The focus was not on how working in a team affects the Cognitive Load of individuals, but whether the concept of Cognitive Load Theory can be applied to a team.

The work of Paul Kirschner and colleagues demonstrated that this is fundamentally possible. However, unlike individual learning, there are additional aspects to consider and necessary adaptations.

The foundation is the hypothesis of a collective working memory within a team that exists only within that team. This collective working memory is formed through communication and coordination among team members. Each person contributes knowledge and helps connect shared understanding. This means new team members cannot automatically access what the team has already learned. The collective working memory thus provides a strong rationale for stable teams.

What’s particularly compelling about the idea of collective working memory is that not every team member needs to learn everything, but individuals can focus on certain aspects. This makes it possible to manage the Intrinsic Cognitive Load of a very complex task by distributing it within the team. However, this doesn’t work for every type of problem: While certain aspects of learning a business domain don’t need to be understood by all team members, concepts related to tools are not easily divisible.

It may work effectively for a team implementing payroll accounting to have two people understand payroll tax and two others understand electronic reporting to health insurance companies. Such a division won’t work with fundamental concepts: If two team members know how to declare variables in the programming language used, and two others know how to work with functions, no one masters the programming language well enough to use it effectively.

Kirschner and colleagues have shown that collective learning in a team works particularly well when the subject is too complex for an individual—meaning the Intrinsic Cognitive Load is too large—and the learning subject is equally new to all team members.

If the team has mixed levels of experience with the learning subject, negative learning outcomes can occur, particularly for team members who already have experience with the subject. These can be explained partly by group biases but are practically unavoidable.

An aspect that plays no role in individual learning has a dramatic effect on group learning: Transactive Activities contribute significantly to Extraneous Cognitive Load. Transactive Activities encompass all communication and coordination efforts that team members must make to learn together. The costs of Transactive Activities can be so high that they negate the advantages of collective learning. It’s therefore worth considering how these Transactive Activities can be reduced to minimize Extraneous Cognitive Load.

Reducing Cognitive Load

Sweller and colleagues have repeatedly demonstrated that high Cognitive Load hinders learning. Reducing Cognitive Load has been a central goal in educational practice for many years.

For individual psychological Cognitive Load Theory, five principles have been identified that lead to a reduction of Germane Cognitive Load, with Extraneous Cognitive Load being the focus:

Coherence: simple, clear instructions and the elimination of superfluous images, sounds, or other stimuli.
Signaling: marking and highlighting the aspect that should be in focus.
Redundancy: non-competing elements with consistent content support learning, e.g., matching audio explanations for an image or labeled elements in a diagram.
Spatial contiguity: related elements are placed close together.
Temporal contiguity: related elements are visible simultaneously.

In addition, there are the didactic methods mentioned earlier to reduce Intrinsic Cognitive Load by meaningfully dividing the subject matter.

These methods naturally also apply to Cognitive Load within teams. However, as described above, additional factors can help reduce Cognitive Load in teams:

Complexity of the learning subject: The task must be complex enough to justify the additional effort of collective learning for all team members.
Guidance and support: Supporting and guiding the team during learning through suitable environment, resources, and methods reduces Transactive Activities.
Domain experience: The more experience team members have in the domain, the fewer Transactive Activities are required.
Collaboration experience: The more experience team members have working together, the fewer Transactive Activities are necessary.
Team size: Smaller teams require fewer Transactive Activities.
Roles: Clearer roles for team members in the learning situation reduce Transactive Activities.
Team composition: More heterogeneous teams require more Transactive Activities.
Work experience: More experienced team members require fewer Transactive Activities with the learning subject.
Previous collaboration: Team members with more experience working together need fewer Transactive Activities. Transactive Activities represent the main driver of Extraneous Cognitive - Load when learning in teams.

Article