Technology-enhanced mathematics learning: review of the interactions between technological attributes and aspects of mathematics education from 2013 to 2022

Mathematical standards (content and process) and technology

While various technologies were employed across different content standards, DMS and CAS emerged as a content-specific technology used exclusively for geometry and algebra, respectively, highlighting their specialized application. Application software, on the other hand, was utilized to deliver a wide range of content domains, and was exclusive to instances where multiple mathematical strands were addressed. For example, in the study by Outhwaite et al. (2020), two mathematic applications Maths 3–5 and Maths 4–6 were used for one-on-one early mathematics instruction. These applications focused on topics such as number, shape, space, and measure, providing structured learning activities designed to support young children’s acquisition of basic mathematical skills.

Notably, none of the studies focused specifically on measurement as a standalone topic. Since number and operations are among the first aspects of mathematics that students learn, and form the foundation for more advanced mathematics, it is understandable that a large proportion of interventions in this area made use of application software. The positive outcomes suggest that applications are the best suited technology for this content. Relatively less effort has been put into tackling more abstract topics such as algebra and geometry, indicating a gap in the research that future studies could fill.

In 13.64% (n = 6) of the studies, the topics covered were not specified, which is an improvement on the review by Crompton and Burke (2015) in which 64% of studies failed to specify the topics covered. These six interventions involved multiple schools and long durations; that is, the technology was not confined to one or a few topics. For instance, Roschelle et al. (2016) used an online tool ASSISTments, a CAA, to support students across 42 schools with homework over an entire school year. ASSISTments provides students with timely feedback and hints while they complete their assignments, and gives teachers organized, real-time data on student performance, allowing for more effective monitoring and intervention. Genlott and Grönlund, (2016) utilized Google Apps for education to engage 502 elementary school students in learning mathematics through communication and discussion over 3 years. Supplemental Table 3 provides a summary of the software aligned to some of the content standards.

Application software proved to be very versatile, suitable for all five process standards, and showed a high correlation with problem solving. Unexpectedly, DMS and CAS were the only technologies not aligned with problem solving. The few studies using DMS and CAS focused on other cognitive activities. A computer algebra system (n = 1, 2.27%) has not yet been developed for standards other than reasoning. Dynamic mathematics environments (n = 3, 6.82%) were often used for representations (Ng et al. 2020; constructing physical solids using 3D printing pens), reasoning and proofs (Kaplar et al. 2022; exploring imitative reasoning and creative mathematics reasoning through an interactive learning environment), and connections (Demitriadou et al. 2020; exploring a real-life three-dimensional world). Although no process standard was entirely connected to any specific type of technology, the opposite is the case; that is, most technology tended to align with only a subset of the standards. For example, the functionality of CAS supports higher-order thinking, which can explain its unique alignment with reasoning and proofs. The interactive and analytical nature of CAI and CAA allow for engagement and collaboration; scaffolding and feedback were a great fit for problem solving.

While problem solving was the most common process identified in the reviewed studies, this finding should be interpreted with caution for three reasons. First, studies may have utilized multiple process standards, but only the most prominent one was selected to avoid double coding. Second, there is evidence of the term “problem solving” still being used in the traditional sense, that is, referring to learners applying procedures to solve well-defined problems (e.g., Lai et al. 2015; Roschelle et al. 2016). For instance, university students used an integrated LMS and CAS to complete algebra exercises with an interactive step-by-step solution (Akugizibwe and Ahn, 2020) and an online application to practice procedural vector skills (Mikula and Heckler, 2017). Third, while some authors claimed to have engaged students in problem solving, lower-level, traditional, and inadequate assessment tools such as tests were utilized rather than more authentic assessment tools such as a mathematical problem-solving behaviors scale (Nye et al. 2018) or a rubric for problem-solving ability (Hsiao et al. 2018). Moreover, TEML studies provided limited explanations and examples of both the types of problems solved and the tools used for measuring learning performance.

In TEML, learners are engaged in a variety of processes to facilitate content learning. Problem solving is used for all seven content standards, with a strong link observed between this process and number and operations. Four of the five processes were utilized for number, operations and geometry. However, perhaps due to the abstract nature of algebra, no attempts have been made to develop and evaluate mathematical conjectures, engage students in coherent discussions about algebraic concepts, or relate these concepts to other domains or the real world. TEML may still be limited to tasks that supplement traditional teaching, failing to capitalize on the transformative potential of technology to facilitate tasks that are inconceivable without its use.

Interaction among technology types, role of technology, and pedagogical approaches

Analysis of the reviewed studies (Supplementary Fig. 5) showed that instruction was the most popular role of technology in both periods. This finding is consistent with previous reviews indicating that TEML interventions were mainly used for instructional enhancement, such as by supplementing regular classes (Rakes et al. 2020). For example, Lai et al. (2015) implemented CAI via video- and game-based materials for remedial math outside regular class hours. Similarly, Vanbecelaere et al. (2020) used embedded curriculum-aligned digital games to support early numerical skills development. Both studies demonstrated how technology applied to instruction enabled alternative ways for students to learn and practice mathematics. Echoing Bray and Tangney’s (2017) findings, most technology-enhanced instruction fell under the augmentation level, serving as substitutes for conventional teaching with functional or conceptual enhancements such as feedback or individualized support.

Our research also revealed that the role of computation was used exclusively in the first period, while visualization and modeling emerged in the later stage. This shift may be attributed to two factors. First, wider availability of technological advancements has led to more enriched instruction. For instance, Augmented Reality and Virtual Reality facilitates students’ 3D exploration, outperforming text-based methods for fostering conceptual understanding (Demitriadou et al. 2020). Similarly, the Interactive Learning Materials Triangle enables students to manipulate shapes, thereby deepening their understanding of geometry (Kaplar et al. 2022). Hoyles (2016) emphasized that such technologies extend visualization and modeling beyond conventional approaches, fostering mathematical creativity and innovative thinking. Second, this trend parallels a pedagogical shift from focusing on content delivery to empowering students to model, analyze, and solve real-world problems (Bakker et al. 2021; Santos-Trigo, 2024). Therefore, TEML has expanded beyond improving computation to cultivating higher-order thinking.

Figure 3 illustrates the relationships among types of technology, their roles, and associated pedagogies. As previously noted, instruction dominates TEML research and is mostly delivered through versatile applications, the cross-platform functionality of which stimulates classroom adoption. Our findings also indicate that technology-assisted instruction is prevalent across various pedagogies, implying that teaching tasks are most readily technologized, compared to other roles. Technologies, such as CAI, ITS, and LMS, facilitate instructional delivery, whereas specific tools and technologies, such as CAA, DMS, and CAS, target narrower functions. Studies involving combination roles integrate multiple technologies. Hence, effective TEML relies on matching tools to their intended pedagogical role (Ran et al. 2021a; 2021b).

From the pedagogy perspective, technology plays more diverse roles—beyond instruction, it supports visualization and combination purposes, which, according to the reviewed studies, were not evident in other pedagogies. This may suggest that, while these technology roles are typically used in guided pedagogy to help learners better understand materials, they also contribute to inquiry-based learning. For instance, Dynamic Geometry Environments effectively fostered inquiry-based learning and advanced students’ geometric understanding (Ng et al. 2020). Conversely, collaborative learning and game-based learning primarily rely on application-type tools for instructional delivery. This pattern reflects both the ubiquity of such applications and their interactive nature or game-oriented format. For instance, application software facilitated interactivity in play-based environments, leading to learning gains and promoting collaboration among kindergarteners (Miller, 2018). It also supported social interactions among learners, peers, and teachers, enhancing mathematics performance (Genlott and Grönlund, 2016). This explains the frequent use of applications in collaborative learning and game-based learning to facilitate instruction.

The observed trends regarding the relations among technology, role, and pedagogy substantiate that certain technologies may have specific affordances. Along with Young’s (2017) three didactical functions of technological interventions, we can conclude that CAS and some application software enable “doing mathematics” or computational efficiency; LMS, ITS, computer-based instruction, and application software are best suited for the development of procedural knowledge and practicing mathematics skills, while DMS affords modelling and exploration for the development of conceptual knowledge. Linking these functionalities to pedagogy, our findings indicate that guided, game-based and collaborative learning are more aligned with computational efficiency and procedural knowledge, whereas inquiry-based learning, characterized by granting students autonomy and responsibility, is better suited to developing conceptual knowledge. Building on this, Engelbrecht and Borba (2024) emphasized that the use of effective pedagogical approaches paired with the right technology can ensure conceptual understanding.

TEML outcomes

The interest in cognitive outcomes, particularly achievement, parallels the findings of Bray and Tangney (2017) and Hwang and Tu (2021). This indicates that many TEML interventions are designed primarily to improve students’ performance on standardized measures of mathematical knowledge. While collaboration and communication were considered as means to acquire content in two studies, they were not outcome variables in any study. This absence may be because these studies primarily aimed to assess the impact of technology on specific mathematical knowledge or skills. For example, Wang et al. (2024) reported significant gains in mathematics achievement through features that supported interaction and knowledge co-construction. In many cases, collaboration serves as a pedagogical strategy to facilitate deeper learning and engagement, rather than as an explicit goal of assessment (Demir and Zengin, 2023). Although some studies in the broader computer-supported collaborative learning literature assess social skills (e.g., Bringula and Atienza, 2023), measuring collaborative skills such as communication or collaborative problem solving requires different methodologies, such as observation or interaction analysis. Consistent with Hussein et al. (2022), the limited attention to assessing collaborative competencies highlights a research gap in understanding how collaborative dynamics influence learning outcomes, with subject matter knowledge remaining the dominant focus. Communication and collaborative skills have gained importance in mathematics, as evidenced by their inclusion in the 2015 PISA assessment framework using computer-based assessments in PISA (OECD, 2017). Engelbrecht and Borba (2024) highlighted the growing relevance of student online collaboration in mathematics education via learning environments and social media. Therefore, future TEML studies may consider communication and collaboration as a cognitive learning outcome.

A more in-depth examination of achievement outcomes suggests that further research is needed to confirm the positive effects of TEML interventions. Studies that had mixed effects utilized multiple measures of achievement (Ng et al. 2020; Vanbecelaere et al. 2020) or a delayed test (Ng et al. 2020). This implies that the effects of technology are tied to the specific content and may not last over an extended period. For instance, Vanbecelaere et al. (2020) revealed that elementary school students who played digital games outperformed the control group engaged in non-gaming instruction on number line estimation, but not on number comparison.

Our findings revealed that engagement in TEML was satisfying and provided a positive learning experience for participants. However, results regarding the other affective dimensions such as attitude, self-efficacy, and motivation were less consistent or depended on the affordances for interaction, adaptability, and innovation, yielding varied results. Fabian et al. (2018) noted that the use of tablets and the SnapShot Bingo application was not enough to outweigh traditional daily mathematics activities or to change elementary school students’ attitudes. The increased accessibility and use of application software in and outside classrooms might eliminate the novelty effect, indicating that merely incorporating technology does not automatically enhance student interest in mathematics. Sustained gains emerge only when the digital tools are embedded in well-designed tasks that allow students to connect to meaningful contexts, and are supported by timely teacher scaffolding and feedback.

The assessment of participants’ self-efficacy in CAI and motivation during the use of application software produced inconsistent results. Immediate evaluative feedback in computer-based training utilizing static question sequences resulted in improved self-efficacy (Heckler and Mikula, 2016). At the same time, computer-assisted tutoring sessions held outside of regular school hours to solve routine problems were found to be ineffective (Lai et al. 2015). A game-based application that lacked vivid animations and scenarios had no distinct effect on motivation (Es-Sajjade and Paas, 2020). In contrast, the use of AR applications to support mobile learning in out-of-school learning environments had a positive effect on motivation (Poçan et al. 2022). These findings suggest that certain technological features, such as vivid animations and immediate feedback, may be better suited to particular affective outcomes like motivation and self-efficacy. Finally, relatively few studies incorporated ITS, DMS, and multiple types of software, although those that did all showed significant affective outcomes. This finding suggests that future research should systematically investigate how adaptive tools such as ITS and DMS can be purposefully aligned with pedagogical strategies to cultivate lasting gains in motivation and self-efficacy and identify which specific features deliver the biggest impact across diverse learner populations.

All instances where emotion and technology acceptance were measured reported non-significant differences. Participating in mathematics professional development did not manifest more frequent use of technology (Thurm and Barzel, 2020). Engaging in gameplay to learn number sense did not alter elementary school students’ mathematics anxiety (Vanbecelaere et al. 2020), and providing a free laptop to partake in a computer-assisted online summer mathematics program was not effective in terms of raising enjoyment (Lynch and Kim, 2017). Extensive research has examined the relationship between emotions and mathematical outcomes. A recent longitudinal study reported positive reciprocal relations between emotions and mathematics performance (St Omer et al. 2023). However, only two TEML studies in the past decade evaluated emotions, and both yielded insignificant results. This necessitates an increased examination of the varying emotions students experience as they engage with technology in their mathematics classes.

To make strides in improving and better understanding the affective domain in TEML, researchers should consider including more observational data and mixed methods of data collection (Grootenboer et al. 2008). For instance, Fabian et al. (2018) used interviews and an activity evaluation in addition to questionnaires to obtain comprehensive information about negative attitudes towards mathematics and to explain the insignificant changes in attitudes. While quantitative metrics can capture overall attitudes and improvements in learning, qualitative data from interviews, observations, and artifacts provide deeper insights into students’ perceptions and experiences of the TEML environment, including their sense of autonomy in gamified conditions (Ortiz-Rojas et al. 2025), cognitive processes, and learning strategies (Demir and Zengin, 2023). The triangulation of multiple sources with performance data enhances the validity and credibility of the study findings.