A Categorization of Workplace Learning Goals for Multi-Stakeholder Recommender Systems: A Systematic Review

The concept of workplace learning refers to a broad range of learning activities through which individuals acquire or improve work-related skills and competencies. These learning activities can occur in formal, non-formal, and informal settings at the physical workplace or in other locations such as at home or educational institutions (Doyle & Young, 2007; Kyndt & Baert, 2013; Manuti et al., 2015). According to Matthews (1999), workplace learning should promote the development and performance of the individual learner as well as the organization. Therefore, workplace learning is determined by the reciprocal interaction between individual and organizational demands and needs to balance the goals of both the individual learner and the organization (Kyndt & Baert, 2013; Manuti et al., 2015). While the initial approaches to workplace learning were primarily classroom-based training, workplace learning is now becoming increasingly digitalized (Ifenthaler, 2018; Manuti et al., 2015). Digital technologies and artificial intelligence enable the implementation of more flexible, cost-effective, and customized forms of workplace learning (De Laat et al., 2020) as well as new methods to support learning processes such as learning analytics (Giacumo & Breman, 2016; Ifenthaler, 2022), adaptive learning (Plass & Pawar, 2020), or educational recommender systems (Tavakoli et al., 2022).

Recommender systems are designed to support users in areas where the amount of information exceeds the individual’s abilities to process it and provide them with the individually most relevant, interesting, or useful object (Manouselis et al., 2012). Classic approaches like collaborative filtering or content-based recommender systems (e.g., Koren et al., 2009) assume that similar users prefer the same objects and are successfully employed in commercial applications and information services. The similarities between the tasks of recommender systems in the consumer domain and the challenges of online education have led to many adaptations, especially for the personalization of learning processes (Drachsler et al., 2015). Examples of successful adaptation include finding relevant learning content (Deschênes, 2020), entire courses (Guruge et al., 2021), or the optimal sequence of learning content and activities (Kerres & Buntins, 2020).

At a second glance, however, recommender systems in the educational domain impose a significantly higher complexity than recommender systems employed in the consumer domain (Drachsler et al., 2007; Kerres & Buntins, 2020). In addition to general interests and preferences, systems must address a variety of other dimensions to sufficiently reflect and support learners' learning processes, for example, continuously changing knowledge and skill levels, varying learning goals, and time constraints. Even for learners with the same interests, different learning activities often need to be recommended depending on their prior knowledge and skill levels, learning goals, and time constraints. Consequently, the use of recommender systems for educational purposes is rather limited so far (Krauss, 2018), especially in the context of workplace learning (Rivera et al., 2018).

We want to address two major drawbacks of existing (educational) recommender systems. Firstly, they are inherently short-term-oriented in the sense that they try to meet the present demands of the user and maximize the current satisfaction (Burke et al., 2011; Manouselis et al., 2012). However, successful learning sometimes demands significant amounts of planning (Margaryan et al., 2013) and requires recommender systems to take a long-term perspective instead. Current advances in the use of reinforcement learning for (commercial) recommender systems put more focus on the maximization of long-term rewards (Afsar et al., 2021). However, they cannot be translated to the workplace learning setting without a proper categorization of workplace learning goals. Educational recommender systems require a proper framework that models learning processes and from which sound recommendations for learning activities and materials can be derived. A categorization of workplace learning goals can help to build such a framework by describing the exact objectives and reasons why learners engage in workplace learning and by enabling transparency and interoperability of recommender systems (Ilkou et al., 2021; Reichow et al., 2022).

Secondly, specifically for workplace learning, current educational recommender systems lack a multi-stakeholder perspective (Abdollahpouri et al., 2020; Burke et al., 2016), that is, they do not put the organization and its corporate strategies into the focus of attention alongside the learner. The strategic and organizational conditions set by the employer (e.g., the selection of content-related learning goals and the learning opportunities, contents, and activities necessary to achieve them) and the setting of the time frame highly influence the learning process and its effectiveness (Doyle & Young, 2007; Kyndt & Baert, 2013; Matthews, 1999). Yet, they are so far widely ignored by educational recommender systems.

Workplace learning goals are cognitive representations of states or outcomes an individual strives to achieve in workplace learning (Deci et al., 1996; Locke & Latham, 2002; Seijts et al., 2013). They are key drivers of learning motivation as they direct attention towards goal-relevant activities and enhance effort and persistence (Locke & Latham, 2002). According to self-determination theory (Deci et al., 1996), goal-directed behaviors differ in their degree of self-regulation. Behavior-based on intrinsic motivation is considered highly self-regulated, because it is performed out of pure interest (Deci & Ryan, 2000; Gagné & Deci, 2005; Vansteenkiste et al., 2006). Goal-directed behaviors that are not intrinsically motivated are based on extrinsic motivation, that is, the behavior is not performed out of pure interest but is related to separable consequences (e.g., rewards or punishments). Extrinsically motivated behavior can be more or less self-regulated depending on the degree to which the individual has internalized the regulation of the behavior and integrated its associated value into the self. If the extrinsically motivated behavior is valuable to the individual, it is considered as (highly) self-regulated. However, if the behavior is controlled exclusively by external demands or contingencies, its self-regulation is very low. For example, if individuals engage in workplace learning because getting a job promotion is important to them, the behavior is extrinsically motivated but self-regulated. However, if individuals engage in workplace learning, because they experience pressure from their supervisors, the degree of self-regulation is very low and the regulation is considered external (Deci & Ryan, 2000; Deci et al., 1996; Gagné & Deci, 2005). Thus, Deci et al. (1996) suggest that self-regulation is a dimension ranging from intrinsic motivation (high self-regulation) to external regulation (low self-regulation).

Following Deci et al.‘s (1996) self-determination theory, we suggest that workplace learning goals can be arranged along a dimension from intrinsic to external goals. Intrinsic goals are related to the individual learner’s interests and are highly self-regulated (Vansteenkiste et al., 2006). In contrast, external goals are not set by the individual learner but by external demands or contingencies usually defined by the organization. External goals are based on the least self-regulated form of extrinsic motivation (Deci et al., 1996; Gagné & Deci, 2005). Thus, intrinsic and external goals can be arranged along Deci et al.’s (1996) self-regulation dimension, with intrinsic goals corresponding to the extreme with the highest degree of self-regulation and external goals corresponding to the extreme with the lowest degree of self-regulation.

We suggest that different workplace learning goals and their associated degree of self-regulation affect how individuals engage in learning. For example, educational research has shown that the degree of self-regulation of learning goals is positively associated with learners’ persistence (Vansteenkiste et al., 2006), achievement (Grolnick et al., 1991), goal attainment, and satisfaction (Skues et al., 2019). Moreover, individuals engaging in workplace learning because they are generally interested in the topic and want to learn something new (= intrinsic goal) might benefit from other learning materials than individuals who need to solve a specific work-related problem defined by external demands or contingencies (= external goal). Thus, recommender systems for workplace learning should be aware of the different workplace learning goals that learners pursue and their degree of self-regulation to identify those learning materials necessary to enhance the development of both the individual learner and the organization (Buder & Schwind, 2012; Gagné & Deci, 2005; Plass & Pawar, 2020).

In educational research, several research efforts have attempted to categorize learning goals in the form of taxonomies (e.g., Bloom et al., 1956; Carter, 1985; Krathwohl, 2002). For example, Bloom’s taxonomy of learning goals distinguishes between cognitive, affective, and psychomotor goals, each of which is subdivided into different hierarchical levels (Bloom et al., 1956; Harrow, 1972; Krathwohl & Bloom, 1964). Moreover, Carter (1985) developed a taxonomy of learning goals for professional education and distinguishes between learning goals related to knowledge (knowing), skills (doing), and personal qualities (being). However, these taxonomies might not be useful for recommender systems for workplace learning as they either do not focus on the context of work or ignore the varying degrees of self-regulation related to workplace learning goals. Therefore, this paper aims to develop a new categorization of workplace learning goals and their degree of self-regulation. For this purpose, we conducted a systematic review guided by the following two research questions:

To present our results, we have chosen the word categorization instead of taxonomy, as the categories of workplace learning goals developed in our systematic review are neither organized hierarchically nor orthogonal to each other as usually supposed by taxonomies (e.g., Bloom et al., 1956).

Our systematic review followed the guidelines proposed by Okoli (2015). To identify the body of research investigating workplace learning goals, we conducted an electronic search in the databases ACM Digital Library, Google Scholar, PsycINFO, and Web of Science. Following the methods of other recently published systematic reviews on workplace learning (e.g., Watson et al., 2018), we set the publication period to one decade (2011–2021). We searched for the following search terms: goal or objective in combination with (and) “professional training” or “professional development” or “vocational training” or “vocational education” or “workplace learning” or “further education” or “continuing education”.

A total number of N = 13,198 publications were exported from the databases. After removing duplicates, the remaining N = 11,923 publications were screened for the following inclusion criteria: The publication (1) presented empirical findings, (2) examined workplace learning goals (we excluded publications that only examined goal attainment without reporting the goal content as well as publications focusing on samples of higher education or K-12 students), (3) was based on a sample with non-disabled adults, (4) was written in English or German, and (5) was published in a peer-reviewed journal. N = 34 key publications met all inclusion criteria and were included in the systematic review. Figure 1 provides an overview of the different steps of the screening process. As publication screening progressed, a saturation of workplace learning goals was reached, that is, the consideration of further publications did not produce additional workplace learning goals. Therefore, we assume that a review of publications from the past decade was sufficient to capture different research trends on workplace learning goals.

We developed a data extraction sheet to extract workplace learning goals and study characteristics (country, research design, sample size, sample characteristics, workplace learning scenario) from the key publications. Deductive and inductive content analysis (Mayring, 2015) was deployed to synthesize results: Based on Deci et al.’s (1996) self-determination theory, we arranged the extracted workplace learning goals along a dimension from intrinsic to external goals and then inductively clustered them into different categories by grouping similar learning goals and aggregating them at a higher level of abstraction. Data extraction and synthesis were conducted and discussed in iterative processes by two of the authors and two trained student assistants.

Educational recommender systems offer benefits for workplace learning by tailoring the selection of learning activities and materials to the individual learner (Drachsler et al., 2015). However, existing educational recommender systems are short-term oriented and focus on the learner as the primary stakeholder of learning processes (Assami et al., 2018; Burke et al., 2011; Laseno & Hendradjaya, 2019). As workplace learning needs to balance the demands and goals of both the individual learner and the organization (Kyndt & Baert, 2013; Matthews, 1999), recommender systems for workplace learning should become more long-term oriented and incorporate the organizational perspective as well (Abdollahpouri et al., 2020; Burke et al., 2016).

This systematic review aimed to develop a categorization of workplace learning goals for multi-stakeholder recommender systems incorporating both the individual learner’s and the organization’s perspectives. We reviewed current publications that investigated workplace learning goals and identified five categories of workplace learning goals. These categories vary in their degree of self-regulation and can be arranged along a dimension from intrinsic goals (high self-regulation, i.e., set exclusively by the learner) to external goals (low self-regulation, i.e., set exclusively by the organization).

When engaging in workplace learning, learners pursue different learning goals such as advancing one’s career or fulfilling mandatory job requirements. This systematic review provides a categorization of workplace learning goals and shows that workplace learning goals range from intrinsic goals (set exclusively by the learner) to external goals (set exclusively by the organization). The categorization of workplace learning goals offers a common language for educational recommender systems incorporating both the individual learner and the organization as major stakeholders. Future research should build on our categorization to develop recommender systems for workplace learning to adapt the behavior of educational recommender systems to the specific types of workplace learning goals.