AI in the Domain
Describing the use of AI in the domain is the starting point of any endeavor to create a domain-specific AI course, as it determines what content will be taught in the corresponding courses. The subtopics of the pillar are presented in the following subsections.
Domain
The term domain is used to refer to the discipline in which AI is to be applied. For AI applications in medicine, an exemplary domain could be “radiology” and for AI applications in vehicle technology and development, it could be “mechanical engineering”.
Potential AI Use Cases in the Domain
This subtopic is focused on the effect of AI technology in the domain. It helps to structure the topics that will become relevant for students and learners in the near future. Its goal is to support the identification of current use cases and the prognostic assessment of possible future use cases in which AI could play a role in the domain-specific problem solution.
Data in the Domain
The identified AI use cases are usually based on the most relevant type of data in the respective domain. The respective subtopic is not about which data is easy to obtain or how it can be used but rather about the type of data that is involved. Knowledge of typical data in a given domain enables more targeted use of AI techniques and specification of the data. It makes a big difference for the AI techniques which are to be taught whether the domain mainly works with time-series data, texts, images or other data types. Moreover, it is an important consideration whether the data in the domain is abundant or scarce.
Implications of Using AI in the Domain
Another important factor to consider is the potential implications that could arise when using AI in the respective field
[32]. This mostly concerns ethical, legal and social implications
[33][34]. For example, using AI to support medical triage decisions has different implications than using it to optimize the energy consumption of a manufacturing plant. Understanding the impact of technology on their domain helps students adhere to societal and ethical standards when using or developing AI technologies in their domain.
Additional Learning Resources
The creation of course material can be supported and guided by existing learning material and from oneself, colleagues or other institutions. In particular, Open Educational Resources (OER) can provide a basis for course development and can be used as preparatory or supplementary materials in course design
[35][36].
Overall, the answers to the questions from this pillar build the foundation on which skills and knowledge are to be taught in the course units.
Learning Environment
In addition to the external aspects of AI in the domain, there are several ways in which the learning environment in which the course takes place can influence the pedagogical implementation of the course. Domain-specific AI courses as interdisciplinary courses place a special demand on learners, instructors and internal support. Thus, it is important to fully understand who the learners are, what skills the instructor possesses and which additional internal support is available.
Learners and Their Interaction with AI
Concerning the learners, three considerations are important for domain-specific AI courses. First, it is important to understand which AI skills and related competencies such as mathematical foundations, computational literacy, data literacy or programming skills the learners have acquired beforehand. Second, it is important to clarify the role of the group of learners regarding their interaction with AI to choose relevant demonstrations of AI-applications and an appropriate level of difficulty. The role can be described in different ways, e.g., using the taxonomy suggested by Faruqe et al.
[37], which presents four groups whose frequency of contact with AI and AI competency requirements differ from each other. According to the authors, the levels in ascending order are “Consumers, the General Public and Policymakers”, “Co-Workers and Users of AI Products”, “Collaborators and AI Implementers” and “Creators of AI”
[37]. Third, the existing competencies and the future role are influenced by the curricular integration of the course in an overall program. Moreover, the curricular integration determines if it is a mandatory course and correspondingly the expected number of students in the course. Note, that depending on the interdisciplinarity of the course, the group of learners may be more heterogeneous with students from different fields and with different experiences.
Instructors
Next to learners, instructors play an important role in the learning process
[38]. Domain-specific AI teaching requires a mix of sufficient AI knowledge, domain expertise and pedagogical skills to teach an interdisciplinary course as well as the motivation and time from an instructor’s perspective. The AI knowledge of faculty and instructors tends to be quite heterogeneous, ranging from no previous AI experience to decades of AI research experience
[11]. Thus, it is important to understand and assess the instructor’s abilities to teach the course. However, if instructors self-assess their knowledge and skills in AI, attention must be paid to possible cognitive biases or heuristics, leading to an under- or overestimation of actual AI skills. In various other professional contexts, it has been found that people relatively rarely assess their skills correctly and it can be assumed that it is no different when assessing one’s own AI knowledge and skills
[39][40].
Internal Support
Internal support, such as budget, personnel restraints, the maximal duration of courses, available data, software and hardware, can be viewed as resources in a positive sense or as limitations in a negative sense. In the context of AI teaching, two important considerations are the availability of data and the availability of hardware and computing resources. Moreover, instructor support (e.g., through training), institutional barriers concerning interdisciplinary teaching and student support (e.g., through additional resources and infrastructure) play a role in designing an interdisciplinary course
[9].
Course Implementation
The right pillar represents the core of the framework as it combines the findings from the previous pillars. It aims to create a pedagogical structure that can be interpreted as a short version of the final course implementation. The pillar is structured following the Constructive Alignment approach
[24], aligning the desired learning outcomes, the assessment of those outcomes and the respective learning activities.
Learning Outcomes
Defining the content and scope of the learning outcomes is an important building block in the context of domain-specific AI teaching that is informed by the considerations of the other pillars and determines the focus of the course. To organize learning outcomes in a structured, consistent and verifiable manner, it is recommended to formulate learning objectives
[24]. Learning objectives should be specific, measurable, achievable, reasonable and time-bound (i.e., “SMART” objectives;
[41]) wherever possible. Furthermore, they should focus on specific competence levels following Bloom’s Taxonomy
[42]. The course learning objectives determine the structure of the course and indicate the time and resources spent on the individual topics. The learning objectives should be shared with the students so that they know which aspects of the course are the most relevant for their professional development.
Assessment
Following the Constructive Alignment approach, it is important to consider in advance through which methods and in which way the fulfillment of the learning objectives will be evaluated
[24]. Assessment in interdisciplinary courses requires to balance the experiences of different groups of learners as well as the targeted outcome with respect to their interaction with AI. In addition to traditional assessment methods such as exams, tests oral presentations or reports, the applied nature of domain-specific AI teaching can also benefit from project- or problem-based assessments that are connected to real-world applications (see
[43] as an example). Moreover, research in interdisciplinary education indicates that using assessment through reflection can help students to bridge the disciplinary silos
[9]. Similar to other fields, using different assessment components can be a beneficial and fair approach to account for the different experiences of students from different disciplines
[44].
Learning Activities
The last step focuses on the learning activities that lead to the desired learning objectives
[24]. Thus, the focus is on the pedagogical implementation of the overall course design. In this context the Merrill principles of learning
[25] should be considered to promote an effective learning experience. Experiences from the few domain-specific AI courses that are already conducted today, indicate that a combination of different teaching methods is often used to address the different aspects of AI
[45]. The overview of the learning activities builds the basis for more detailed planning of the learning activities throughout the course. These could also include using AI-based learning activities.
Intended Use of the AI Course Design Planning Framework
The AI Course Design Planning Framework forms a visual and practical tool for instructors and course developers in the higher education or professional education context with a special focus on non-computer science (non-CS) students. It can be used as means to gather ideas, innovate, plan and communicate ideas for domain-specific AI courses. The framework can be used as a self-contained instrument for individuals, in tandem with AI and domain experts or in a workshop setting with multiple people. The authors suggest filling it from left to right, first considering the questions on AI in the domain, the learning environment of the course and last the course implementation.