Introduction

Product anthropomorphism, the practice of attributing human characteristics to products to create a more vivid and relatable product image by triggering human cognitive processes, has become a widespread strategy in marketing (Kim et al., 2016, Mueller et al., 2017, Huang et al., 2020). Marketers utilize anthropomorphism in various forms, such as incorporating humanized mascots or anthropomorphic animals in advertisements (e.g., M&M candies, Three Squirrels, Want Want), designing products with human-like features (e.g., the Coca-Cola bottle or Jean Paul Gaultier’s perfume bottles shaped like the human body), or adding subtle human cues—like eyes or mouths—into product visuals.

The rise of e-commerce platforms has significantly contributed to the design and growing popularity of anthropomorphic products. However, this approach also carries risks (Hoffman and Novak, 2018, Dennis, 2024). Numerous studies have examined the potential negative impacts of anthropomorphism on consumer relationships, particularly in the context of adverse events such as brand or service failures (Puzakova et al., 2013, Choi et al., 2021). One line of research specifically addresses the risks associated with chatbot anthropomorphism during service failures, highlighting how it can result in consumer dissatisfaction and even anger (Crolic et al., 2022). Another stream of research has focused on product failures, specifically the anthropomorphism of brands and foods (Puzakova et al., 2013, Schroll, 2023). A common hypothesis within these work is that anthropomorphism elevates consumer expectations regarding a product’s agency and performance, thereby intensifying disappointment when those expectations are unmet (Crolic et al., 2022).

Moreover, anthropomorphic features often cause consumers to attribute failures directly to the brand or product (Puzakova et al., 2013). As a result, much of the existing research on product or service failure has concentrated on understanding the negative effects of anthropomorphism from the perspective of consumer cognition, such as perceived capacity for pain (Schroll, 2023). However, there is limited exploration of implicit factors such as emotional responses (Pavone et al., 2022, Verhulst et al., 2025). While consumers may rationally assess a product failure based on observable criteria, their underlying emotional reactions are often swift, automatic (Walker, 2019), and profoundly influential in shaping their overall satisfaction and future behavior (Crolic et al., 2022). For instance, even when consumers recognize that a failure is not entirely due to the product’s inherent qualities, the human-like features that initially fostered a connection may still trigger strong, unconscious emotional responses that intensify negative evaluations. These issues motivate the current exploration. Given that emotions are important in decision-making and can significantly shape consumer behavior (Andrade, 2005, Han et al., 2007, Lerner et al., 2015), a deeper understanding of these emotional reactions is necessary to comprehend the effects of product anthropomorphism in marketing contexts.

It is noteworthy that an individual’s physiological responses—such as emotion, attention, and cognition—can be influenced by external cues like anthropomorphic features (Yin et al., 2024, Han et al., 2023). In high-stakes situations, such as product failures, these implicit emotional processes can determine whether a consumer ultimately forgives a brand or severs their relationship with it (Tsarenko and Tojib, 2012). Thus, analyzing the negative effects of anthropomorphism through a more comprehensive understanding of consumer responses can enhance our insight into the effective use of anthropomorphism in e-commerce. With the growing role of social media, the social influence of product failure has become increasingly significant. Nevertheless, due to a lack of effective measurement methods, many previous studies relying on traditional self-report techniques have focused only on consumer’s explicit responses (e.g., consciously initiated cognition and emotion) (Fox et al., 2018), neglecting their unconscious emotional reactions. Consequently, this paper aims to address the following question: How does anthropomorphism affect consumer satisfaction in the context of product failure, and what role do emotions play?

To address the research question, we conducted both a survey and a laboratory experiment using event-related potentials (ERPs). Study 1 initially validate the interaction effect between product effect and anthropomorphism, showing that consumers are less satisfied with anthropomorphic products compared to their non-anthropomorphic counterparts, particularly in the context of product failure. Study 2 further explored the emotional responses driving this effect. Given the difficulty of measuring emotions directly through self-report methods (Gregor et al., 2014, Kuan et al., 2014), Study 2 employed ERPs to capture participants’ brain responses. Such method can objectively identify emotional reactions as consumers process product information. Our results indicate that in product failure, anthropomorphism activates early automatic emotional processing, marked by increased P2 amplitudes. This, in turn, leads to prolonged disappointment in subsequent stages of processing. The convergence of these emotional responses ultimately renders anthropomorphic products less satisfying than non-anthropomorphic ones. This study underscores the pivotal role emotions play in consumer dissatisfaction with anthropomorphic products in failure contexts.

Our work advances two research streams. First, we supplement existing literature on product anthropomorphism by introducing implicit consumer emotions into the assessment of satisfaction in product failure scenarios. Previous studies have primarily focused on cognitive aspects, such as perceived capacity for pain (Schroll, 2023), with limited exploration of implicit emotions—an essential factor in understanding the formation of decisions (Westbrook and Oliver, 1991). Second, this research adds to anthropomorphism theory by providing a deeper understanding of how consumers process negative information about anthropomorphic products. We show that both early automatic cognitive violations and later sustained negative emotional responses are implicit mechanisms that lead to greater consumer dissatisfaction following the failure of anthropomorphic products. Moreover, our findings offer practical guidance for retailers, suggesting that managing consumer expectations may mitigate the prolonged negative emotional impact of product anthropomorphism. In the following sections, we review key literature, develop our hypotheses, present two empirical studies, and discuss our findings and contributions.

Literature review

Our work is related to three streams of literature, including the impact of anthropomorphism, the expectancy violation theory (EVT), and the role of emotion. Here, we discuss each of these strands of research and describe how we contribute to them.

Product anthropomorphism and product failure

Anthropomorphism, the attribution of human characteristics to non-human entities, has attracted considerable attention in fields such as chatbots (Araujo, 2018), service robots (Kim et al., 2019), and product or brand anthropomorphism (Hur et al., 2015), among others. In consumer research, anthropomorphic representations lead consumers to perceive brands as living entities with human-like motivations, traits, conscious will, emotions, and intentions (Kim and McGill, 2011). Product anthropomorphism specifically refers to the attribution of human-like physical and mental characteristics to inanimate objects (Landwehr et al., 2011, Aggarwal and McGill, 2007, Huang et al., 2020). This strategy is considered effective because anthropomorphism is believed to induce positive changes in consumer attitudes toward the product (Puzakova et al., 2013). For instance, Chandler and Schwarz (2010) found that consumers who anthropomorphize their cars are less inclined to replace them, irrespective of the cars’ quality, emphasizing traits similar to those found in human relationships. Likewise, Chen et al. (2017) demonstrated that consumers facing threats to their sense of belonging are motivated to form social connections with brands that display human-like characteristics.

While the positive aspects of anthropomorphism have been well-documented (Chen et al., 2023, Aggarwal and McGill, 2007), numerous studies have examined its negative impacts, particularly in the context of product or brand failure (Puzakova et al., 2013, Choi et al., 2021). For instance, when anthropomorphic chatbots fail to meet performance expectations, they can exacerbate consumer dissatisfaction and frustration (Crolic et al., 2022). Likewise, anthropomorphic packaging in food can intensify feelings of betrayal and disappointment when the product does not meet consumer expectations (Aggarwal and McGill, 2012).

However, much of the research on anthropomorphism in failure scenarios has focused on how it raises customers’ expectations about a chatbot’s agency and performance capabilities, leading to expectancy violations when performance falls short (Crolic et al., 2022). Previous studies have rarely explored how implicit emotions, particularly those that occur after expectation violation, influence consumer satisfaction with anthropomorphic products in failure scenarios. This research explores the emotions that drive the negative effects of anthropomorphism in product failure and their impact on satisfaction.

Expectancy violation theory

EVT is a widely used theory about individual perception, which was first proposed by Burgoon and Jones (1976). It provides a way to explain how individuals perceive and interpret violations of their personal space. According to EVT, when an expectation violation occurs, individuals initially devote increased attention to the aspect of the interaction that has been violated. Then, they process and respond to these expectation violations through a sequential process of interpretation and evaluation. The evaluation phase enables individuals to better understand the violator, the nature of the violation, and how to respond (Bachman and Guerrero, 2006).

Recently, EVT has been expanded to marketing contexts to understand how consumers evaluate the products or services’ performance and compare that to their pre-usage expectations (Cadotte et al., 1987). Expectancy violations not only damage satisfaction (Oliver, 1980, Oliver and Swan, 1989) but also have negative effects on customers’ attitudes toward the company and their purchase intentions (Oliver, 1980). More importantly, many studies on anthropomorphism, particularly those examining chatbots, have explored the impact of anthropomorphic features on consumer satisfaction from the perspective of expectancy violation (Crolic et al., 2022, Cai et al., 2024). Therefore, draws on EVT, the current study to examine how anthropomorphism in product failure scenarios influences consumer emotions and subsequently impacts consumer satisfaction.

Consumer Emotion and ERPs in this study

Emotions involved in decision-making are implicit and challenging to measure (Gregor et al., 2014, Kuan et al., 2014). ERP technology is widely employed to detect relevant neural signals with high temporal resolution, enabling the evaluation of the time course of emotional processing (Luck, 2014, Yu et al., 2022). Prior ERP research in marketing has demonstrated that ERP components are precise and effective for capturing individuals’ attentional, expectational, and emotional responses, with particular emphasis on the P2 and LPP components (Pozharliev et al., 2015). We utilize these two components to explore participants’ emotional responses in the context of product failure.

The P2 is an early positive potential peaking around 200 ms after stimulus presentation, with a broad scalp distribution (González-Villar et al., 2014). Previous research has frequently highlighted the P2’s role in reflecting unconscious attentional engagement (Carretié et al., 2001). The LPP is characterized by a slow positive ERP component over the centro-parietal sites, peaking around 500–700 ms after stimulus onset (Pozharliev et al., 2015). The LPP amplitude is widely recognized as a key indicator of attentional engagement with emotional stimuli (Schupp et al., 2000, Zhang et al., 2019), with higher amplitudes occurring more frequently for emotionally significant (both pleasant and unpleasant) stimuli compared to neutral visual stimuli (Olofsson et al., 2008).

Conceptual framework

As discussed previously, product failure serves as a significant emotional stimulus for consumers, with its negative effects potentially spilling over into their evaluations of both the product and the brand (Darke et al., 2010, Valentini et al., 2020, Smith and Bolton, 2002). As noted by Jean Harrison‐Walker (2012), service failure elicits negative emotions such as anger, offense, and disappointment. For anthropomorphic products, anthropomorphism naturally activates human schemas, causing consumers to attribute human-like emotions and intentions to these products (Aggarwal and McGill, 2007), which in turn raises their attention before use. Consequently, we expect to observe a higher P2 amplitude in anthropomorphic products as it is associated with increased initial attentional engagement. Therefore, we first hypothesize that:

H1: Anthropomorphism increases consumers’ attentional engagement before product use relative to non-anthropomorphic products (as shown by the greater P2 amplitudes).

This heightened attentional engagement is closely linked to consumers’ pre-usage expectations. After using the product, consumers assess its performance and compare it to their pre-usage expectations. If the anthropomorphic product fails, it may lead to expectation violations. Although the most immediate negative reaction may appear as reduced product satisfaction, it is also likely to evoke feelings of disappointment on an emotional level (Saeed et al., 2024). Align with H1, we anticipate that, the expectation violation triggers a surge in attention when the product fails, resulting in greater P2 amplitudes, a pattern not observed in the non-failure scenarios. Furthermore, we also anticipate observing a larger LPP amplitude for anthropomorphic products, as they evoke greater affective significance than the non-anthropomorphic ones during the late stages of emotional processing in product failure scenarios. Hence, we propose:

H2: When the product fails, anthropomorphism (vs. non-anthropomorphism) has a negative impact on consumer satisfaction.

H3: When the product fails, anthropomorphism (vs. non-anthropomorphism) elicits stronger emotional responses (as shown by the larger P2 and LPP amplitudes).

Overview

Across two studies, combining self-reported and physiological data, we test our hypotheses. Our investigation began with an online survey (Study 1) to examine whether anthropomorphism intensifies customer dissatisfaction when products fail. Then, Study 2 employs an ERP experiment to capture participants’ emotions while processing product information, exploring how consumers process negative information about anthropomorphic products. Through these two studies, we confirm our hypotheses and conclude by discussing the theoretical and managerial implications.

Study 1

The goal of Study 1 was to examine whether the effect of product anthropomorphism on consumer satisfaction depends on product effectiveness. A 2 × 2 between-subjects design was used, with product type (anthropomorphism vs. non-anthropomorphism) and product effect (poor vs. good).

Participants, procedure, and measures

Participants. A total of valid 337 participants (Mage = 32.26, SD = 9.64; 179 females, details see Table 1) recruited from Credamo (https://www.credamo.com) took part in the study. The criterion for selecting participants was that they have online shopping experience. This ensured that participants could understand our research context. Participants read a detailed informed consent form and agreed to participate before completing an online questionnaire (see Ethics declarations for details). In addition, participants were required to be fully engaged and attentive during the experimental tasks; therefore, we included attention-filter question. Thirteen responses were excluded due to incorrect answers to an attention-filter question. All participants had prior online shopping experience and were randomly assigned to one of four conditions.

Table 1 Descriptive statistics of the samples in study 1 (N = 337).
Full size table

Materials. To ensure the product’s broad applicability, Study 1 used commonly encountered Bluetooth speakers as stimuli, manipulating the level of anthropomorphism in these speakers (see Fig. 1). The stimuli were adapted from (Qin et al., 2024), with the anthropomorphism level altered through the inclusion of facial features (including eyes, ears, and mouth). Other aspects of the speakers were largely unchanged to prevent distracting participants with variations in unrelated product features, such as trademarks or colors. Additionally, the stimuli featured a classic black color, commonly found in electronic products.

Fig. 1: Materials and results in study 1.
Fig. 1: Materials and results in study 1.
Full size image

a Illustration of experimental materials; b Satisfaction results. Error bars: ±1 SE; **p < 0.01, ns p < 0.05.

Procedure and measures. Initially, participants were instructed to imagine purchasing a Bluetooth speaker online, with the product being displayed in either anthropomorphic or non-anthropomorphic forms. Participants were then told that they had purchased the speaker and experienced either poor or excellent sound quality during use. They were then asked to rate their satisfaction with the purchase. Finally, their demographic information was collected. The measurement items are shown in Table 2. The results show that the standard factor loadings for all items exceed the 0.5 threshold recommended by Hair et al. (2019), and both composite reliability (CR) and average variance extracted (AVE) meet the standards proposed by Fornell and Larcker (1981), indicating that the scale exhibits good reliability and convergent validity.

Table 2 Measurement Items in Study 1.
Full size table

Results

Manipulation Check. Prior to group comparisons, the normality and homogeneity of variance assumptions were assessed via the Kolmogorov–Smirnov test and Levene’s test, respectively. Both assumptions were violated: non-anthropomorphic group (D = 0.13, p < 0.001) and anthropomorphic group (D = 0.09, p = 0.001) deviated significantly from normality, while Levene’s test indicated unequal variances (F = 4.83, p = 0.03). Therefore, parametric tests such as the independent-samples t-test were deemed inappropriate. To address this, the Mann-Whitney U test was selected for group comparisons. Results revealed a statistically significant difference in anthropomorphism ratings between the non-anthropomorphic (Mdn = 2.80, IQR = 1.80) and anthropomorphic conditions (Mdn = 4.20, IQR = 2.40), with the latter receiving higher scores (U = 8330, p < 0.001, r = 0.36). The significant difference between conditions validates the experimental manipulation.

Consumer satisfaction. Prior to analysis, the assumptions of normality and homogeneity of variances were assessed. The Kolmogorov–Smirnov test revealed significant deviations from normality across all groups (e.g., Poor Effect-Non-Anthropomorphic: D = 0.25, p < 0.001; Good Effect-Anthropomorphic: D = 0.20, p < 0.001). Levene’s test further indicated unequal variances (F3, 333 = 5.08, p = 0.002). Given these violations, the Aligned Rank Transform (ART) method was employed to conduct a non-parametric two-factor ANOVA (Wobbrock et al., 2011). The ART analysis indicated that the main effect of product effect was significant (F1, 333 = 689.22, p < 0.001, η2p = 0.67). Consumer satisfaction was notably lower in the poor effect condition (MPoor = 2.08, SD = 1.27) compared to the good effect condition (MGood = 5.84, SD = 0.77). The main effect of product type was not significant (F1, 333 = 3.20, p = 0.07). There was a significant interaction between product type and product effect (F1, 333 = 4.07, p = 0.04, η2p = 0.01). Further post hoc analysis revealed that the influence of product type varied under different effect conditions. Using Estimated Marginal Means (EMM) and Tukey-adjusted pairwise comparisons, significant differences between groups were identified. In the Good-effect condition, product anthropomorphism did not significantly influence satisfaction (p = 0.97). However, in the Poor-effect condition, anthropomorphic products (MAnthropomorphic = 1.80, SD = 0.97) led to lower satisfaction than non-anthropomorphic products (Mnon-Anthropomorphic = 2.35, SD = 1.46; p = 0.005).

Discussion

Study 1 confirms the hypothesis related to the negative effect of anthropomorphism (H2): when a product fails, anthropomorphism has a more negative impact on consumer satisfaction compared to non-anthropomorphism. In Study 2, we aim to explore the underlying mechanisms behind this effect, specifically examining the role of emotions in the decision-making process.

Study 2

Study 2 seeks to explore how participants process negative information about anthropomorphic products and how their emotional responses influence satisfaction evaluations, thus providing support for H1 and H3. Given that emotions are often subtle and difficult to observe directly, Study 2 employs ERPs to record participants’ brain activity, a method increasingly utilized in management research (Kuan et al., 2014). Additionally, Study 2 employs a 2 × 2 within-subject design (product type x product effect) and uses an S1–S2 paradigm to simulate product failure scenarios, enabling a natural examination of the interaction effect between these two factors.

Participants

Fifty-eight right-handed, healthy undergraduate and graduate students were recruited from a large public university in China. The criteria for selecting valid participants in Study 2 were consistent with those in Study 1. Six participants were excluded due to excessive artifacts during electroencephalogram (EEG) recordings, resulting in a final sample of 52 participants (26 males, Mage = 20.78, SD = 3.17, details see Table 3)Footnote 1. All participants had normal or corrected-to-normal vision and no history of neurological or mental illness. Participants were compensated with 60 RMB (approximately 10.00 USD) for their participation, with an additional performance-based incentive of up to 30 RMB (approximately 5.00 USD). Informed consent was obtained at the beginning of the study (see Ethics declarations for details).

Table 3 Descriptive Statistics of the Samples in Study 2 (N = 52).
Full size table

Materials and procedure

The S1–S2 experimental paradigm was used, where two stimuli were presented sequentially (Yang et al., 2013). The first stimulus (S1) displayed product images, either anthropomorphic or non-anthropomorphic. Forty images were selected from a pool of items commonly needed by college studentsFootnote 2. Anthropomorphism refers to attributing human traits, emotions, and intentions to non-human entities (Epley et al., 2007). Products without explicit anthropomorphic features, such as eyes and mouths, were shown as non-anthropomorphic. In contrast, these features were added for the anthropomorphic versions. The images were standardized in shape, size, and resolution (270 × 360 pixels), with unnecessary details removed using Adobe Photoshop 2019 (Adobe Systems Incorporated, San Jose, CA, USA). Following this design principle, two versions were created for each product, resulting in a total of 80 images (See Fig. 2b). The second stimulus (S2) presented Chinese words indicating the product effect, either good or poor (See Fig. 2c). Both types of stimuli were randomly paired, resulting in 160 trials. These trials were evenly divided into four blocks, each containing 40 trials.

Fig. 2: Procedure and materials in Study 2.
Fig. 2: Procedure and materials in Study 2.
Full size image

a Example of the procedure for a single trial. b, c S1 and S2 materials.

In this task, participants imagined an online shopping experience and then rated their satisfaction with the product after receiving it and learning about its effects. As shown in Fig. 2a, each trial started with a black cross (+) on a light gray background for 600–800 ms. The product images were then displayed for 1000 ms, followed by a blank screen for 1000 ms. Next, the words representing the product’s effect were presented for 1500 ms. Finally, participants rated their shopping satisfaction on a scale from −3 (very dissatisfied) to 3 (very satisfied), using the provided keyboard. The specific item was “Please indicate your satisfaction with this shopping experience” adapted from Van Ittersum et al. (2013). Participants were instructed to press the “1” key for lower satisfaction scores and the “3” key for higher scores, confirming their choices by pressing “2”. The interval between blocks was 2 min. After completing all 160 trials, participants provided demographic information.

EEG data acquisition, preprocessing and analysis

EEG data acquisition

Participants were seated in a soundproof room, approximately 100 cm from a monitor displaying stimuli, after shampooing and blow-drying. They were then fitted with EEG caps, and conductive gel was applied (Fu et al., 2022). To ensure the quality of EEG data, participants were instructed to minimize blinking, eye and muscle movements. A 10-trial practice session was conducted before the formal experiment. EEG data were recorded using a 32-channel elastic cap (Brain Products GmbH, Gilching, Germany) with Ag/AgCl electrodes placed according to the international 10–20 system (Sharbrough, 1991), with a band-pass filter of 0.1–100 Hz and a sampling rate of 500 Hz. The FCz electrode was used as the reference during recording, and impedances were kept below 10 kΩ. We used E-Prime 3.0 software to present stimuli and record triggers and responses (Psychology Software Tools, Pittsburgh, PA, USA).

EEG data analysis

EEG data preprocess. EEG data were analyzed using the EEGLAB v14.1.1 (Delorme and Makeig, 2004), running on MATLAB R2021b (The Mathworks Inc.). The average of the left and right mastoids (TP9 and TP10) was used as the offline reference. A digital low-pass filter at 30 Hz was then applied to the continuous EEG signals. Next, EEG recordings from −200 to 800 ms time-locked to the onset of S1 and S2, were extracted for ERP analysis. Specifically, P2 analysis during the product type evaluation stage covered the period from 200 ms before S1 stimulus to 800 ms after S1, with the first 200 ms of the pre-stimulus interval used as the baseline. Similarly, P2 and LPP analyses during the outcome feedback stage covered the period from 200 ms before S2 to 800 ms after S2, also using a 200 ms pre-stimulus baseline. Each participant’s EEG data was then visually inspected to discard any poor-quality epochs. Independent component analysis (ICA) was applied to reduce artifacts from eye blinks, eye movements, and muscle activity. Finally, trials with amplifier clippings, significant electromyographic noise, or peak-to-peak fluctuations exceeding ±100 µV were excluded to ensure data quality for subsequent analysis. Figure 3 presents the protocol for EEG data acquisition, preprocessing, and analysis.

Fig. 3
Fig. 3
Full size image

Protocol of EEG data acquisition, preprocessing and analysis.

ERP analysis. The EEG data collected during the product type evaluation stage were divided into two categories: non-anthropomorphic and anthropomorphic conditions. For the outcome feedback stage, the data were further classified into four conditions: non-anthropomorphic products with good or poor effects, and anthropomorphic products with good or poor effects. Following visual inspection and the guidelines proposed by Picton et al. (2000), we analyzed the P2 component during the product type evaluation stage, and both the P2 and LPP components during the outcome feedback stage. The time windows and electrode sites were as follows: (1) product type evaluation stage: P2 from 160 to 220 ms after S1 onset at P3, P4, Pz, O1, and O2; (2) outcome feedback stage: P2 from 200 to 280 ms and LPP from 500 to 700 ms after S2 onset at P3, P4, Pz, O1 and O2Footnote 3 (Pozharliev et al., 2015).

Results

Behavioral results

A two-way repeated-measures ANOVA was conducted. Results revealed significant main effects of anthropomorphism type (F1, 51 = 10.565, p = 0.002, η2p = 0.172), product effect (F1, 51 = 391.441, p < 0.001, η2p = 0.885), and a significant interaction between anthropomorphism type and product effect (F1, 51 = 14.714, p < 0.001, η2p = 0.224). Participants rated their satisfaction with non-anthropomorphic products (Mnon-AM = 0.016, SD = 0.261) higher than with anthropomorphic products (MAM = −0.070, SD = 0.275). Additionally, participants rated their satisfaction with good-effect products (Mgood = 1.476, SD = 0.695) higher than with poor-effect products (Mpoor = −1.531, SD = 0.494). A simple effects analysis of the interaction showed that anthropomorphism (vs. non-anthropomorphism) significantly reduced satisfaction for products with poor effects (Mpoor_non-AM = −1.446, SD = 0.501; Mpoor_AM = −1.615, SD = 0.508; F1, 51 = 33.343, p < 0.001, η2p = 0.395), but had no significant effect on products with good effects (F1, 51 = 0.004, p = 0.952, η2p = 0.000). These effects are visualized in Fig. 4.

Fig. 4: Behavioral Results of Study 2.
Fig. 4: Behavioral Results of Study 2.
Full size image

a Averaged satisfaction rating in non-AM and AM conditions. b Averaged satisfaction rating in good-effect and poor-effect conditions. c Averaged satisfaction rating under four conditions. Error bars = ±1 SE. *p < 0.05, **p < 0.01, ***p < 0.001; ns means not significant. AM means anthropomorphism.

ERP results

P2 component at the product type evaluation stage. We conducted a 2 (product type: anthropomorphism vs. non-anthropomorphism) × 5 (electrode: P3/Pz/P4/O1/O2) repeated-measures ANOVA on the P2 amplitude in the 160–220 ms time window. The results revealed a significant main effect of product type (F1, 51 = 18.364, p < 0.001, η2p = 0.265) and electrode (F1, 51 = 22.871, p < 0.001, η2p = 0.656). There was no significant interaction effect between product type and electrode (F4, 204 = 1.619, p = 0.185, η2p = 0.119). These results suggested that the anthropomorphism condition elicited a larger P2 amplitude (MAM = 5.412 µV, SD = 3.057) than the non-anthropomorphism condition (Mnon-AM = 4.354 µV, SD = 2.808) (see Fig. 5).

Fig. 5: P2 results at the product type evaluation stage.
Fig. 5: P2 results at the product type evaluation stage.
Full size image

a Grand averaged P2 waveforms for non-AM and AM conditions from the Pz electrode; b Topographic maps of P2 for non-AM and AM. The time window is 160–220 ms. c Averaged P2 amplitudes comparison between non-AM and AM conditions. Error bars = ±1 SE, ***p < 0.001. AM means anthropomorphism.

P2 component at the outcome feedback stage. We conducted a 2 (product type: anthropomorphism vs. non-anthropomorphism) × 2 (product effect: good vs. poor) × 5 (electrode: P3/P4/Pz/O1/O2) repeated-measures ANOVA for P2 amplitudes. The main effect of product type was insignificant (F1, 51 = 1.339, p = 0.253, η2p = 0.026), and the product effect was also insignificant (F1, 51 = 3.119, p = 0.083, η2p = 0.058), but a significant interaction effect was found (F1, 51 = 6.011, p = 0.018, η2p = 0.105). The three-way interaction effect among product type, product effect, and electrode was also non-significant (F4, 204 = 0.915, p = 0.463, η2p = 0.071). Thus, we focused on the interaction effect between product type and product effect. Further simple effects analysis indicated that P2 amplitude was significantly greater in the anthropomorphism condition than in the non-anthropomorphism condition for products with poor effects (Mpoor_non-AM = 4.578, SD = 3.406; Mpoor_AM = 5.553, SD = 3.318; F1, 51 = 4.368, p = 0.042, η2p = 0.079), while no such effect was observed for products with good effects (F1, 51 = 1.052, p = 0.310, η2p = 0.020) (see Fig. 6).

Fig. 6: P2 results at the outcome feedback stage.
Fig. 6: P2 results at the outcome feedback stage.
Full size image

a Grand averaged P2 waveforms from the Pz electrode, elicited by viewing products (AM, non-AM) and product effects (Good, Poor). b Topographic maps of P2 for the difference between product-type (AM minus non-AM) waves for both Good (left) and Poor (right) conditions within the interval of 200–280 ms. c Averaged P2 amplitudes comparison under these four conditions. Error bars = ±1 SE, *p < 0.05, ns not significant. AM means anthropomorphism.

LPP component at the outcome feedback stage. We conducted a 2 (product type: anthropomorphism vs. non-anthropomorphism) × 2 (product effect: good vs. poor) × 5 (electrode: P3/P4/Pz/O1/O2) ANOVA for LPP amplitude in the 500–700 ms time window. The results revealed no significant main effects for product type (F1, 51 = 0.522, p = 0.473, η2p = 0.010) or product effect (F1, 51 = 0.058, p = 0.811, η2p = 0.001), but a significant interaction effect was observed (F1, 51 = 7.258, p = 0.010, η2p = 0.125). The three-way interaction effect among product type, product effect, and electrode was also non-significant (F4, 204 = 2.0, p = 0.11, η2p = 0.143). Similarly, we then focused on the interaction effect between product type and product effect. Further simple effects analysis showed that LPP amplitude was significantly greater in the anthropomorphism condition than in the non-anthropomorphism condition for products with poor effects (Mpoor_non-AM = 2.294, SD = 4.005; Mpoor_AM = 3.517, SD = 3.933; F1, 51 = 4.299, p = 0.043, η2p = 0.078), while no such effect was observed for products with good effects (F1, 51 = 1.633, p = 0.207, η2p = 0.031). See Fig. 7 for details. Table 4 summarizes the ERPs results.

Fig. 7: LPP Results at the Outcome Feedback Stage.
Fig. 7: LPP Results at the Outcome Feedback Stage.
Full size image

a Grand averaged LPP waveforms from the Pz electrode, elicited by viewing products (AM, non-AM) and product effects (Good, Poor). b Topographic maps of LPP for the difference between product-type (AM minus non-AM) waves for both Good (left) and Poor (right) conditions within the interval of 500–700 ms. c Averaged LPP amplitudes comparison under these four conditions. Error bars = ±1 SE, *p < 0.05, ns not significant. AM means anthropomorphism.

Table 4 Summary of the ERP Results.
Full size table

Discussion

Study 2 directly supports H1 and H3 from an electrophysiological standpoint. Using ERP technology, we found that anthropomorphism initially triggers early automatic emotional processing, marked by an increased allocation of attentional resources. This attentional investment intensified when participants experienced product failure, resulting in prolonged emotional processing during later stages. The integration of these emotional responses ultimately renders anthropomorphic products less satisfying than non-anthropomorphic products when failures occur.

General discussion and conclusion

Previous research has examined consumers’ responses to anthropomorphic and non-anthropomorphic products in the context of product or service failures. However, the emotional processing of anthropomorphic products during product failure remains underexplored, despite its importance for understanding consumer perceptions of such products. We used two studies to examine how anthropomorphism affects consumer satisfaction in the context of product failure, as well as the role of emotions in this process. Study 1 offers initial evidence that anthropomorphism, compared to non-anthropomorphism, reduces consumer satisfaction in cases of product failure. Study 2 uses an ERP experiment to examine emotional responses during product information processing, revealing the underlying mechanisms of this phenomenon. The ERP results indicate that anthropomorphic features trigger consumers’ automatic emotional processing, leading to the allocation of greater attentional resources to these products. However, in the context of product failure, these anthropomorphic features cause sustained disappointment during later stages of processing, making anthropomorphic products less satisfying. In the next sections, we will discuss the theoretical and managerial implications of these findings.

Theoretical implications

This research makes several significant theoretical contributions to the field of anthropomorphic product marketing and consumer behavior literature. First, it expands the literature on product anthropomorphism by examining the formation of consumer satisfaction through implicit emotional processing in the context of product failure. Previous studies have primarily focused on consumers’ cognitive responses, such as perceived capacity for pain (Schroll, 2023), or self-reported emotional reactions. However, emotional processing is inherently implicit, complex, and difficult to capture accurately (Gregor et al., 2014, Kuan et al., 2014), yet it is essential for understanding the development of satisfaction. Our research advances existing literature by revealing how emotional factors, at the physiological level, influence the development of product satisfaction. By integrating these implicit emotional responses, we extend the scope of product anthropomorphism research (Huang et al., 2020, Qin et al., 2024), shedding light on the interplay between consumer emotions and product perceptions, and highlighting the need for a more nuanced approach that acknowledges the complexity of emotional processing in the context of product interactions.

Second, this research advances the theory of anthropomorphism by deepening our understanding of its negative effects in the context of product failure. Unlike previous studies that often treat consumer reactions as monolithic (Walker, 2019), our combined findings from Experiments 1 and 2 reveal that the negative impact of product anthropomorphism unfolds as a continuous process. Crucially, we used ERP components (P2 and LPP) to capture different stages of emotional processing during consumer decision-making. The results suggest that early automatic cognitive violations and later sustained negative emotional arousal are implicit processes, explaining why consumers show greater dissatisfaction with anthropomorphic products after a failure. For providers of anthropomorphic products, failing to take effective measures to prevent product failures may lead to more negative consumer attitudes than with non-anthropomorphic products. In conclusion, this research provides a new perspective on the negative effects of product anthropomorphism, emphasizing the importance of addressing these implicit emotional processes.

Managerial implications

The findings of this research offer valuable managerial implications for companies that incorporate anthropomorphism into product design and marketing. Although anthropomorphic features can attract attention and foster emotional connections (Yao et al., 2023), our work reveals that these features may also lead to greater disappointment when failures occur. Managers therefore should carefully evaluate the risks of anthropomorphizing products, particularly in cases where failures are likely or difficult to prevent. To minimize the negative effects of product failures on consumer satisfaction, companies should enforce strict quality control and implement effective failure management strategies. For anthropomorphic products, managing consumer expectations by clearly communicating product capabilities and limitations can reduce the likelihood of expectation violations. For example, when launching anthropomorphic products, companies can offer clear service guarantee commitments or specify product limitations (Crisafulli and Singh, 2016) to reduce the impact of potential failures. Finally, the study also emphasizes the importance of understanding consumers’ emotional reactions to anthropomorphic products. In the event of a failure, providing personalized and empathetic customer service early in the emotional response may help reduce dissatisfaction and maintain brand loyalty timely. Moreover, Companies can utilize emotion analytics to gain insights into these implicit responses, refining design and communication strategies to better align with consumer expectations and emotional needs.

Limitations and future research

While our work provides valuable insights, it also has certain limitations. Study 2 involved a relatively small and homogeneous sample, primarily composed of college students. Although this group represents a key segment of online consumers, their responses to e-commerce information are generally reflective of broader populations, and their high compliance makes them suitable for EEG experiments, the limited diversity in the sample may restrict the generalizability of our findings. Future research should incorporate more diverse samples to enhance the applicability of the results across different demographic groups. Moreover, this study focused on immediate emotional responses; future studies could take a longitudinal approach to investigate how these emotions impact long-term behaviors, such as brand loyalty. Finally, this study focuses on explaining the emotional effects of anthropomorphism in product failure. While we examined key demographic factors (e.g., age, gender, education level), the influence of cultural variables warrants attention in an increasingly globalized world, as prior research indicates varying consumer attitudes toward anthropomorphism across cultural contexts (Baskentli et al., 2023, Mehmood et al., 2024). We will take this into consideration in our future study.