Introduction

In the last decade, emotional design, which refers to the use of design features to promote positive emotions (Norman, 2007) or pleasure in users (Jordan, 2000), has attracted the interest of researchers in many fields (Brom et al., 2018; Cao et al., 2021; Pengnate & Sarathy, 2017; Roy & Naidoo, 2021; Song et al., 2021; Triberti et al., 2017). In multimedia learning settings, emotional design cues (e.g., human-like features and pleasant colors) have been employed to important elements of learning materials (Heidig et al., 2015; Mayer & Estrella, 2014). Such manipulations have the potential to induce positive emotions in learners, attract learners’ attention, and enhance learning outcomes (Brom et al., 2018; Mayer & Estrella, 2014; Schneider et al., 2018).

Facial anthropomorphism refers to adding facial elements such as eyes and mouth to non-human graphical elements (Brom et al., 2018). It is assumed that facial expressions can communicate emotions (Ekman & Rosenberg, 1997). Moreover, the baby-face bias suggests that things with baby-faced features are more likely to induce positively-valence reactions (Brom et al., 2018). In educational settings, Mayer and Estrella (2014) employed anthropomorphism design principles in a multimedia lesson on how a virus causes a cold by rendering the host cell and virus each as a round and symmetrical face with eyes; results indicated that the anthropomorphism design improved students’ retention. In contrast, other studies did not find that facial anthropomorphic features improved learning outcomes compared to neutral design (Liew et al., 2022; Shangguan et al., 2020b; Slabbert et al., 2022). For example, Liew et al. (2022) examined the effects of anthropomorphic design in learning materials that explain on how a distributed denial-of-service attack occurs (i.e., adding eyes and mouths on the images); results indicated that anthropomorphism influenced learners’ affective motivational states but did not affect learners’ intrinsic motivation and learning outcomes. Slabbert et al. (2022) indicated that anthropomorphized graphics (i.e., those with human-like expressions) that are decorative in nature may not contribute to learning.

Given the mixed findings, the effectiveness of anthropomorphic design requires more supporting evidence. In 2018, Brom et al. (2018) conducted a systematic review to examine the effects of emotional design (including the use of positive colors and anthropomorphism). The results of the meta-analysis indicated that emotional design had a significant positive effect on several learning and affective-motivational outcomes. Wong and Adesope (2021) replicated Brom’s review (2018) by including newly published articles. However, the two reviews used the umbrella term “emotional design” rather than “anthropomorphism design,” which may have masked some important differences among design elements (e.g., using pleasant colors versus anthropomorphism). Whether anthropomorphism in multimedia learning materials influences the learning process independently has not been examined explicitly. Moreover, previous reviews have mainly discussed the effects of emotional design on learning and affective-motivational outcomes, leaving other relevant outcomes unexamined. Considering these knowledge gaps, this study systematically reviewed the existing evidence on the effectiveness of facial anthropomorphism design of multimedia learning materials, assessing whether it improved learning and other relevant outcomes. In addition, we examined how the effects of anthropomorphism differed across learners’ ages and types of learning materials.

METHODS

Search strategy

The review was conducted in accordance with the Cochrane Collaboration guidelines for systematic review (Higgins et al., 2019). A systematic literature search was conducted for studies published before February 2023 using the databases Web of Science, PsycInfo, and PsycArticle. The following search terms were used for the search: (anthropomorphi*) and (effect* or impact or examin* or evaluat* or assess* or compar*) and (learning or comprehension or recall or memory or attention or transfer or retention or motivation or effort).

Study selection

Studies were included in the review if they met the following criteria: (1) were based on experimental design, (2) examined the effects of facial anthropomorphism on learning and other relevant outcomes in the context of multimedia learning; (3) reported quantitative data (e.g., means and SDs for the outcomes) for computing standardized effect sizes; and (4) were published in English-language, peer-reviewed journals. The titles and abstracts of the citations identified in the literature search were read to determine their relevance. The full texts of the relevant articles were then reviewed for final inclusion. The reference lists of the studies chosen for inclusion, as well as those of relevant review articles, were also screened to capture any missed articles. The screening of studies for eligibility was performed by KL and PS independently. Any discrepancies were resolved through discussions until a consensus was reached.

Data extraction

The two authors independently extracted the following characteristics from each study included in the review: author(s), publication year, study location, sample size, mean age of the participants, sex ratio, learning materials, anthropomorphism manipulation, and the statistics of learning/performance outcomes. For studies used a two-way factorial design with anthropomorphism and one secondary factor as independent variables, each level of the secondary factor was considered as independent sample. For instance, the study by Shangguan et al. (2020a, b) used a 2 × 2 between-group design, with learners’ prior knowledge (high vs. low) and anthropomorphism design (presence vs. absence) serving as independent variables. We treated participants with high or low levels of prior knowledge as two independent samples, which resulted in two sets of comparisons (i.e., high level prior knowledge + anthropomorphism design vs. high level prior knowledge + neutral design; low level prior knowledge + anthropomorphism design vs. low level prior knowledge + neutral design). For studies with multiple intervention groups relevant to anthropomorphism, we split the “shared” control group into two or more groups to form two or more pairwise comparisons.

Outcome measures

Outcomes were divided into five categories, including learning outcomes (e.g., retention, transfer), affective-motivational outcomes (e.g., positive affect, intrinsic motivation), effort outcomes (e.g., mental effort), attention outcomes (e.g., fixation duration on specific area of interest), and experience outcomes (e.g., satisfaction, aesthetics).

Data analysis

Outcomes were meta-analyzed if they were reported in at least five trials. We pooled data across trials using random effects models and calculated the standardized mean difference (SMD) for each outcome. Heterogeneity was examined using the I2 statistics, with the values of 25%, 50%, and 75% indicating low, moderate, and high levels of heterogeneity, respectively (Higgins et al., 2003). Publication bias was assessed using the Egger test, with a p value smaller than 0.05 considered as the existence of publication bias (Egger et al., 1997). Subgroup analysis was performed for learning outcomes by learners’ age and type of learning material using Comprehensive Meta-Analysis software. Narrative synthesis was performed for outcomes that were reported in less than five trials.

RESULTS

Study selection

Figure 1 presents the process of literature search and study selection. We identified 33 independent experiments from 13 research articles (Liew et al., 2022; Mayer & Estrella, 2014; Park et al., 2015; Plass et al., 2014; Schneider et al., 2018, 2019; Shangguan et al., 2020a; Shangguan, Wang, Shangguan et al., 2020a, b; Slabbert et al., 2022; Stárková et al., 2019; Um et al., 2012; Uzun & Yıldırım, 2018; Wang et al., 2023).

Study characteristics

Table 1 presents the study characteristics of the 33 independent experiments. Twenty-three (69.7%) were conducted after 2016. More than half of the experiments were conducted in Europe. The median sample size for the 33 experiments was 55 (range, 37 to 90). Table 2 presents the details of the learning materials.

Fig. 1
figure 1

Flow diagram of the study selection process

Full size image
Table 1 Summary of the characteristics of the 33 independent experiments
Full size table
Table 2 Details of the learning materials
Full size table

Meta-analysis results

The meta-analysis results for learning outcomes are presented in Table 3. Forest plots for transfer, retention, and comprehension are presented in Figs. 2 and 3, and 4, respectively. Facial anthropomorphism designs of multimedia learning materials had significant positive effects on transfer (standardized mean difference [SMD] 0.28, 95% confidence interval [CI] 0.15 to 0.40, p < 0.001), retention (SMD 0.31, 95% CI 0.14 to 0.48, p < 0.001), and comprehension (SMD 0.46, 95% CI 0.27 to 0.64, p < 0.001).

Table 3 The meta-analysis results for learning outcomes
Full size table
Fig. 2
figure 2

Forest plot for transfer

Full size image
Fig. 3
figure 3

Forest plot for retention

Full size image
Fig. 4
figure 4

Forest plot for comprehension

Full size image

Table 4 presents the meta-analysis results for the affective-motivational, effort, and experience outcomes. Among the affective-motivational outcomes, the facial anthropomorphism design of multimedia learning materials had significantly positive effects on positive affect, intrinsic motivation, and intrinsic motivation change. There were no significant between-group differences for positive affect change, valence, and valence change. Among the effort outcomes, anthropomorphism had significant positive effects on perceived difficulty, intrinsic cognitive load, and germane cognitive load. It had significant negative effects on extraneous cognitive load and no significant effect on mental effort. Among the experience outcomes, anthropomorphism had significant positive effects on perceived learning, aesthetics, and enjoyment. No significant difference was observed between the groups for satisfaction.

Table 4 The meta-analysis results for affective-motivational, effort, and experience outcomes
Full size table

Subgroup analysis by age and material type on learning outcomes

The analyses of study outcomes by age and material type (e.g., static material: PowerPoint slide or webpage; dynamic materials: video or animation) were performed on learning outcomes (Table 5). Anthropomorphic design improved retention significantly in juveniles but did not have a significant effect on retention in adults, although it improved transfer significantly in both adults and juveniles. Anthropomorphic design also significantly improved the retention of knowledge from static but not dynamic learning materials. Additionally, it improved transfer significantly in both the dynamic and static groups.

Table 5 Subgroup analysis for learning outcomes by age and material type, respectively
Full size table

Narrative synthesis results

A total of 19 outcomes were narratively synthesized (Table 6). There were significant differences between the anthropomorphism group and the control group for task-irrelevant thinking and some of the eye movement measures. Schneider et al. (2018) indicated that the anthropomorphism group scored higher on task-irrelevant thinking than the control group. Park et al. (2015) reported that fixation was longer on relevant pictures and anthropomorphic elements. Starkova et al. (2019) indicated that pictures attracted more attention during initial observation (2s) in the anthropomorphic condition. There were no significant differences between the anthropomorphic and control groups for the other outcomes.

Table 6 Narrative synthesis results
Full size table

Discussion

Main findings

This systematic review identified a total of 33 independent experiments that examined the effects of facial anthropomorphism design in multimedia learning settings. Overall, the meta-analysis showed that the employment of anthropomorphism design achieved better learning outcomes. However, evidence for the improvement of affective-motivational, effort, and experience outcomes was less conclusive. Some of the outcomes favored anthropomorphism and others were not influenced or even negatively affected by anthropomorphic design. Mixed results were also observed in the narrative synthesis.

Consistent with Brom et al. (2018), we reported small to medium effect sizes for transfer (SMD = 0.28), retention (SMD = 0.31), and comprehension (SMD = 0.46). Such results indicated that employing facial anthropomorphism design in learning materials was effective in increasing achievement scores (Uzun & Yıldırım, 2018). According to the cognitive affective theory of learning with media (CATLM) (Moreno, 2006; Moreno & Mayer, 2007), the reason for such improvements may be that appealing learning materials motivated learners to engage in appropriate cognitive processing during learning. Incorporating human-like features to anthropomorphize the learning elements also elevated learners’ the positive affect (SMD = 0.48). However, the anthropomorphism group did not lead to a significantly larger increase of positive affect than the control group. Non-significant differences in valence and valence change were not observed between groups. The results for effort outcomes were inconclusive. Learners who studied the anthropomorphically designed materials perceived the materials as less difficult and reported lower levels of intrinsic cognitive load and higher levels of intrinsic motivation and germane cognitive load (i.e., perceived understanding), congruently with previous meta-analyses of emotional design (Brom et al., 2018; Wong & Adesope, 2021). Previous literature also indicated that such manipulations would require less effort to process and would make learners feel that materials were easy to learn (Salomon, 1984; Tractinsky et al., 2000). However, extraneous cognitive load was significantly higher when anthropomorphic features were added to the learning materials, indicating that learners may have perceived the instructions and explanations used in anthropomorphic design as less clear than instructions associated with materials designed differently. The added anthropomorphic features were complex, which may have distracted and overloaded students, but their positive effects were strong enough to overshadow the additional load (Schneider et al., 2019). The learners in the anthropomorphic group reported higher levels of perceived learning, aesthetics, and enjoyment.

In the subgroup analysis, we found that anthropomorphism design yielded significant improvements in retention only among learners younger than 18 years old. Facial features may have increased the concreteness of the learning materials for juveniles, resulting in better recall performance (Plass et al., 2014; Tse & Altarriba, 2009). Brom et al. (2018) also found that the effects of emotional design on intrinsic motivation were stronger for younger children than for college students. In our analysis of learning materials, facial anthropomorphism significantly improved learners’ retention when static learning materials were used for study. Learners find static learning materials more difficult than animations or videos and consider that learning from them requires more mental effort (Höffler & Leutner, 2007). Although we found that anthropomorphism facilitated the learning process more in static learning materials than in dynamic learning materials, this result was not in line with previous reviews (Brom et al., 2018; Wong & Adesope, 2021) that found there was no significant difference in retention between the two groups. Further research on the relationship between the material type and learning outcomes is needed.

Our narrative synthesis indicated that anthropomorphic designs significantly influenced task-irrelevant thinking. However, only Schneider et al. (2018) have reported such a result previously, probably because the anthropomorphic features employed in the experiments were complex and therefore induced more irrelevant thoughts, especially for learners with little prior knowledge. Studies have also shown that anthropomorphism has functioned as a process of empathy (Airenti, 2015; Schneider et al., 2018). For eye movement measures, Park et al. (2015) and Starkova et al. (2019) found that anthropomorphic design is attention-arousing and more likely than non-anthropomorphic design to lead to deep information processing. There were no significant differences between their anthropomorphism and control groups for the other outcomes. Given the mixed results, more research is needed to further understand how they are affected by anthropomorphism.

Implications for research

Our review suggests several implications for research. First, most of the experiments examined only the effects of anthropomorphic design on learning and affective-motivational outcomes; few have assessed effort and experience outcomes. Further experiments should focus more on these outcomes to obtain better understanding of whether, how, and to what extent anthropomorphism influences the learning process. Second, most of the experiments measured affective-motivational and effort outcomes through self-reported surveys, making it difficult to distinguish among different affects and/or load types (De Jong, 2010; Kalyuga, 2011). Future studies may consider assessing learners’ affective-motivational and effort outcomes using more objective methods (e.g., ECG and EEG). Third, our systematic review found that few studies have examined the attention-capturing effects of anthropomorphic design on learners (Park et al., 2015; Stárková et al., 2019; Wang et al., 2023). Further examinations of eye-tracking data are required to better clarify how anthropomorphic features influence learners’ attention distribution. Also, future experiments should quantify the level of anthropomorphism so that its optimal level can be determined. Fourth, most of the studies examined only immediate learning outcomes. However, delayed learning outcomes, which reflect the extent to which learners can remember and comprehend knowledge after an interval of time, are important to investigate as well. Finally, several factors might have moderated the effects of facial anthropomorphism on the learning process. For example, most of the studies we reviewed assessed the use of anthropomorphism in learning materials for children and/or younger adults. How anthropomorphism affects adults/older adults has received less attention. The effects of learners’ cultural backgrounds, familiarity with anthropomorphism design, and prior knowledge on the effects of anthropomorphism also remains unclear (Liew et al., 2022).

Implications for practice

Positive emotion has been regarded as a crucial factor influencing the learning process and outcomes (Liu et al., 2023; Tyng et al., 2017). Our review indicated that facial anthropomorphism design can foster positive emotions among learners and improve learning outcomes. Multimedia designers aiming to improve learners’ transfer, retention, and comprehension are encouraged to consider appropriate ways to add facial anthropomorphism to essential elements of the learning materials (Wong & Adesope, 2021). Iterative human factors evaluations are recommended for designers to confirm the optimal degree of anthropomorphism so that learning materials will enhance learning motivation and positive emotions without adding too much cognitive burden.

Limitations

Our systematic review has several limitations. First, the effects of facial anthropomorphism need to be interpreted with caution due to variability in the design of learning materials. For instance, some of the studies we reviewed used only round, face-like shapes in the learning materials, whereas others included eyes, mouth, nose, and eyebrows. In addition, we did not quantify the level of anthropomorphism, so we cannot recommend an optimal level of anthropomorphism for educational materials. Second, the sample available for meta-analysis was relatively small. Nineteen outcomes (e.g., eye tracking metrics, experience outcomes) had to be analyzed using narrative synthesis, by simply measuring their statistical significance. As more research in the field is published, a meta-analysis could be conducted to investigate the effect size and significant level of these outcomes. Third, publication bias was detected for positive affect, intrinsic motivation, intrinsic motivation change, mental effort, and extraneous cognitive load. The results need to be interpreted with caution for these outcomes. Fourth, we included only articles published in English, which could have led to language and publication bias.

Conclusion

This systematic review examined the effects of facial anthropomorphism on the learning process in multimedia learning settings. Our findings indicate that the facial anthropomorphism design of multimedia learning materials can induce positive emotions in learners and improve their intrinsic motivation, facilitating transfer, retention, and comprehension performance. The use of facial anthropomorphism appears to be more beneficial for learners younger than 18 years old and in the design of static rather than dynamic learning materials. The findings of this study can guide educators and multimedia designers in applying facial anthropomorphism to learning materials to facilitate learning outcomes.