Latent profiles of cross-cultural adaptation among African international students in China: integrating identity recognition, social avoidance, cultural intelligence, and perceived discrimination Authors: Wenda Zhu, Yanhua Zhao Published: 2026 ABSTRACT BackgroundInternational students from high cultural-distance backgrounds face unique adaptation challenges, yet research has predominantly examined these challenges using variable-centered approaches that assume population homogeneity. This study adopts a person-centered approach to identify distinct latent profiles among African international students in China based on identity recognition, social avoidance, cultural intelligence, and perceived discrimination, and examines how these profiles differ in cross-cultural adaptation outcomes.MethodsA mixed-methods research design was employed. Quantitative data were collected from 524 African international students at universities in four Chinese cities (Beijing, Guangzhou, Wuhan, and Wuxi) through questionnaire surveys. Latent profile analysis (LPA) was conducted to identify distinct subgroups, followed by auxiliary analyses comparing profiles on cross-cultural adaptation outcomes. Semi-structured in-depth interviews were conducted with a subsample of 18 participants, with qualitative insights extracted using thematic analysis to contextualize the quantitative findings.ResultsA three-profile solution provided the best fit to the data (entropy = .86, BIC = 5389.78). Profile 1, “Resilient Integrators” (35.5%), was characterized by high identity recognition and cultural intelligence, low social avoidance, and low to moderate perceived discrimination. Profile 2, “Anxious Withdrawers” (21.4%), exhibited low identity recognition and cultural intelligence, high social avoidance and perceived discrimination. Profile 3, “Cautious Navigators” (43.1%), showed moderate levels across all indicators but with notable within-profile variability. The three profiles differed significantly in cross-cultural adaptation (χ2 = 156.82, p < .001), with Resilient Integrators reporting the highest adaptation and Anxious Withdrawers the lowest. However, the difference between Resilient Integrators and Cautious Navigators was smaller than expected, suggesting that moderate resources can yield reasonable adaptation in structured contexts. Chinese language proficiency (OR = 1.74, p < .001) and social support (OR = 1.58, p < .001) were significantly associated with membership in the Resilient Integrators profile. Qualitative interviews (N = 18) largely corroborated the quantitative profiles but also revealed important within-profile heterogeneity, particularly among Cautious Navigators, some of whom described experiences overlapping with the Anxious Withdrawers pattern.ConclusionAfrican international students in China form qualitatively distinct subgroups with different configurations of psychological and behavioral characteristics. Findings provide preliminary evidence that support services tailored to the characteristics of distinct subgroups, such as intensive language training and structured social opportunities, may warrant further investigation as alternatives to uniform support programs.Supplementary InformationThe online version contains supplementary material available at 10.1186/s40359-026-04698-w. Introduction International students undergo complex psychological and behavioral adjustments when adapting to new cultural environments. While much research has examined adaptation among students from culturally similar backgrounds, those from high cultural-distance contexts face distinct challenges that remain understudied. Students from Sub-Saharan Africa studying in East Asia represent one such population, as they navigate substantial cultural distance, language barriers, and visible minority status. Understanding their adaptation mechanisms has important theoretical and practical implications as South-South educational mobility continues to expand [50]. Students from Sub-Saharan Africa encounter adaptation challenges stemming from multiple sources, including differences in communication styles, educational norms, and social customs, the difficulty of learning a tonal language with a non-Latin script, and visible phenotypic differences that may heighten social salience and self-consciousness in cross-cultural interactions. This research gap limits both our understanding of universal patterns in cross-cultural adaptation and the development of targeted support services. Cross-cultural adaptation is a complex process involving psychological, sociocultural, and academic dimensions, where international students must continuously adjust themselves to achieve functional integration in a new environment [6, 10, 62]. Social interaction is widely recognized as a core pathway for cultural learning and adaptation. Through contact with members of the mainstream society, international students acquire cultural norms, establish social support networks, and alleviate culture shock [51, 68]. However, field observations and preliminary interviews reveal a paradoxical phenomenon: a considerable proportion of African students exhibit significant social avoidance tendencies, confining their social circles to co-national or co-ethnic groups, actively reducing contact with Chinese students, and displaying withdrawal behaviors in public settings [12]. This social avoidance behavior appears to be a self-protection strategy that can temporarily alleviate anxiety and discomfort from cross-cultural contact, helping individuals maintain psychological balance [76]. In the long term, however, social avoidance may reduce opportunities for cultural learning, limit the accumulation of social capital, and hinder language proficiency development, potentially contributing to a pattern of avoidance, isolation, and further avoidance [48]. Although social avoidance has been discussed in stress coping literature, the factors associated with it and its relationship with adaptation outcomes in cross-cultural contexts have not been systematically explored. To understand the formation of social avoidance tendencies, it is necessary to consider ethnic identity processes. In this study, identity recognition refers specifically to the degree of ethnic identity exploration and commitment as conceptualized by Phinney [57], that is, the extent to which individuals have actively examined and developed a clear sense of their ethnic group membership. This construct is distinct from broader notions such as bicultural identity integration, host-society identification, or general identity security, which are not measured in the present study. In cross-cultural contexts, international students face dual identity challenges: they need to maintain identification with their home culture while repositioning themselves in a new cultural environment [20, 27]. Phinney’s [57, 58] ethnic identity development model clarifies that stable ethnic identity recognition can provide individuals with psychological anchoring, enhance self-worth, and buffer cultural adaptation stress. For African international students in China, this process is particularly complex because they must maintain their cultural identity while integrating into Chinese society. Drawing on social identity theory [31], individuals who have achieved a high degree of ethnic identity exploration and commitment may experience greater certainty about their group membership and self-worth. This internal sense of security may be associated with greater willingness to participate in cross-group social interactions with a more open attitude [2]. Conversely, individuals with lower levels of ethnic identity exploration and commitment may experience greater uncertainty about group belonging, which in turn may be associated with heightened anxiety and increased reliance on avoidance strategies in cross-cultural encounters [42]. Beyond identity recognition and social avoidance, cultural intelligence (CQ) has emerged as a critical factor in cross-cultural adaptation. CQ refers to an individual's capability to function effectively in culturally diverse settings, encompassing metacognitive, cognitive, motivational, and behavioral dimensions [7, 55]. Research has consistently demonstrated that higher CQ is associated with better cross-cultural adaptation, improved intercultural communication, and greater psychological well-being among international students [13, 21]. For African students in China, CQ may be particularly relevant given the substantial cultural distance they must navigate. Students with high CQ may be better equipped to decode unfamiliar social cues, adjust their behavioral repertoire, and maintain motivation despite cross-cultural challenges [72]. However, CQ does not operate in isolation; it interacts with identity processes, social behavioral tendencies, and environmental stressors to shape overall adaptation patterns. Perceived discrimination represents another crucial factor influencing the adaptation of visually distinctive international student populations. Perceived discrimination refers to the subjective experience of being treated unfairly or differently based on one's group membership, including race, ethnicity, or nationality [28]. For African international students in China, perceived discrimination may manifest through various channels, including interpersonal microaggressions, social exclusion, staring or avoidance behaviors from host nationals, and institutional barriers [54]. Research has established that perceived discrimination is negatively associated with psychological adjustment, life satisfaction, and sense of belonging among international students [41, 64]. Moreover, perceived discrimination has been linked to increased social withdrawal and avoidance behaviors, as individuals seek to protect themselves from further negative experiences [70]. The rejection-identification model suggests that discrimination experiences can trigger a complex process wherein individuals may either strengthen in-group identification as a coping mechanism or experience identity confusion that undermines psychological stability [66]. The four constructs reviewed above, namely identity recognition, social avoidance, cultural intelligence, and perceived discrimination, span different psychological levels. Identity recognition reflects an internal resource rooted in ethnic group exploration and commitment. Social avoidance represents a behavioral tendency in cross-cultural encounters. Cultural intelligence constitutes a multidimensional competence for navigating cultural differences. Perceived discrimination captures subjective appraisal of environmental treatment. It is worth noting that perceived discrimination differs from the other three indicators in that it reflects an appraisal of external conditions rather than an internal psychological attribute. Including it as a profile indicator is intended to capture how individuals' psychological and behavioral characteristics are embedded within their experienced environmental contexts, consistent with ecological and transactional models of adaptation [14, 39]. In variable-centered research, these constructs are typically examined as elements in causal or mediational chains. However, this sequential framing may not adequately represent the lived experience of cross-cultural adaptation, in which these processes operate simultaneously and shape one another in ways that differ across individuals. A growing body of research supports the view that adaptation-relevant processes are better understood as co-occurring configurations rather than as isolated variables arranged in fixed causal sequences. Ward [71] has argued that cognitive, affective, and behavioral dimensions of adaptation operate as an interconnected system rather than as separable predictors. Recent intergroup research further underscores this configural perspective. Borinca, Guerra, and Uka [15] demonstrated that identity processes and belonging needs jointly shape how individuals respond to social inclusion and exclusion, suggesting that identity and social behavioral tendencies are experientially intertwined rather than sequentially ordered. Similarly, Borinca, McAuliffe, and Nightingale [16] showed that responses to cross-group contact are contingent on prior negative experiences and mediated by intergroup affect, indicating that environmental appraisals such as perceived discrimination and behavioral responses such as social avoidance may be best understood as elements of an integrated configuration. Chirkov [22] further argued that acculturation involves simultaneous negotiation of identity maintenance, skill acquisition, and environmental navigation, rather than a stepwise progression through discrete stages. These theoretical perspectives provide a rationale for examining identity recognition, social avoidance, cultural intelligence, and perceived discrimination as co-occurring profile indicators within a person-centered framework. The analytic question shifts from whether one variable causes another to how these constructs naturally combine within individuals to form recognizable subgroup configurations. This configural approach does not preclude the validity of variable-centered causal models. Rather, it addresses a complementary question that has received limited attention in cross-cultural adaptation research with this population. Despite rich achievements in cross-cultural adaptation research, a critical methodological limitation pervades the field: the predominant reliance on variable-centered approaches. Such methods examine average relationships between variables across an entire sample, implicitly assuming that these relationships are homogeneous across all individuals [32]. Cross-cultural psychologists have increasingly called for methodological approaches that move beyond comparing group means toward identifying the processes and configurations underlying cultural variation [46]. Person-centered approaches, particularly latent profile analysis (LPA), offer a methodological solution by identifying unobserved subgroups within a population based on individuals' patterns across multiple continuous indicator variables [8]. In the cross-cultural adaptation literature, person-centered approaches have been increasingly adopted to identify distinct acculturation profiles [29, 33] and cultural identity configurations. However, no study has examined latent profiles integrating identity recognition, social avoidance, cultural intelligence, and perceived discrimination among African international students, a population facing particularly complex adaptation challenges. In summary, three key gaps exist in the current literature. First, there is a notable geographic bias in research samples, with the majority of studies focusing on international students in Western host countries or on East Asian student populations [60]. African students studying in East Asia remain underrepresented. Second, existing research has predominantly employed variable-centered approaches that overlook the possibility that international students form qualitatively distinct subgroups. Third, no study has adopted a configural perspective to examine how identity recognition, social avoidance, cultural intelligence, and perceived discrimination co-occur within individuals to form distinct adaptation-relevant subgroups among this population, despite theoretical grounds for expecting that these constructs operate as interconnected elements of the adaptation experience. The present study addresses these gaps by employing LPA to identify distinct subgroups of African international students in China based on their patterns across these four indicators. Using a mixed-methods design, this study pursues three research questions: RQ1: What distinct latent profiles emerge among African international students in China based on their patterns of identity recognition, social avoidance, cultural intelligence, and perceived discrimination? RQ2: Do the identified latent profiles differ significantly in cross-cultural adaptation outcomes? RQ3: What demographic and contextual factors are associated with membership in specific latent profiles? Methods Research design This study employs a mixed-methods design [5], with quantitative research as the primary component, supplemented by qualitative research for deepened understanding. The rationale is that cross-cultural adaptation is a multilevel phenomenon involving individual psychology, social interaction, and environmental factors. A single method cannot comprehensively capture its richness and dynamics [36]. Quantitative research identifies latent profiles through large-sample questionnaire surveys, establishing the existence of distinct subgroups and their differential adaptation outcomes. Qualitative research explores the subjective experiences and cultural meanings through in-depth interviews, providing experiential context for interpreting the statistical patterns. The two methods complement and cross-validate each other, consistent with the integration principle of mixed-methods research [24, 45]. The mixed-methods approach was essential for three reasons. First, while LPA can identify subgroups, it cannot reveal why and how students develop particular configurations of psychological characteristics. Second, qualitative data serve as a validity check: if the statistically derived profiles are not reflected in participants' narratives, this signals potential measurement or theoretical problems. Third, the experiences of African students in China involve sensitive issues (e.g., perceived discrimination, racial stereotyping) that are difficult to capture through scales alone. Quantitative research Sample and procedure The quantitative research adopted a combination of convenience sampling and snowball sampling strategies, recruiting participants from universities in four Chinese cities: Beijing, Guangzhou, Wuhan, and Wuxi. These cities were selected to represent different geographic regions (north, south, central, and east China) and varying sizes of African international student populations, enhancing the geographic representativeness of the sample. Inclusion criteria were: (a) origin from a Sub-Saharan African country, (b) enrollment in a Chinese university for at least six months, (c) age 18 years or older, and (d) voluntary consent to participate. A total of 524 valid questionnaires were obtained. The sample comprised 312 males (59.5%) and 212 females (40.5%). Regarding age distribution, 378 participants (72.1%) were aged 18 to 25 years, 108 (20.6%) were aged 26 to 30 years, and 38 (7.3%) were aged 31 years or older. City distribution: 156 from Beijing (29.8%), 138 from Guangzhou (26.3%), 124 from Wuhan (23.7%), and 106 from Wuxi (20.2%). Country of origin: 118 from Nigeria (22.5%), 76 from Kenya (14.5%), 68 from Ghana (13.0%), 54 from Ethiopia (10.3%), 47 from Tanzania (9.0%), 32 from South Africa (6.1%), 28 from Uganda (5.3%), and 101 from other countries (19.3%). Major distribution: 168 in engineering (32.1%), 126 in business (24.0%), 98 in medicine (18.7%), and 132 in other fields (25.2%). Time in China: 178 for 6 to 12 months (34.0%), 214 for 1 to 2 years (40.8%), and 132 for over 2 years (25.2%). Data collection took place from September 2025 to December 2025. The study protocol was approved by the institutional ethics committee of Wuxi University Of Technology, with the approved protocol covering multi-site data collection across the four target cities. Data collection procedures adhered to ethical guidelines. The survey adopted a combination of online and offline methods. Online, electronic questionnaires were distributed through the Wenjuanxing platform; offline, paper questionnaires were distributed at international student centers and dormitory areas. All participants were informed of the research purpose, data confidentiality principles, and right to withdraw before participation, and provided informed consent. The questionnaire completion time was approximately 20 to 25 min. Upon completion, participants received compensation equivalent to 20 RMB. To ensure data quality, three attention check items were embedded (e.g., “Please select strongly agree”). Questionnaires with two or more failed attention checks were excluded. Regarding sample size adequacy for LPA, simulation studies suggest that samples of 300 or more can reliably identify 2 to 5 latent profiles across a range of conditions, including varying class separation and profile size ratios [65, 67]. The present sample of 524 participants exceeds recommended minimums and provides adequate statistical power for the three-profile solution identified in this study. Measurement tools Identity recognition The Multigroup Ethnic Identity Measure (MEIM; [57]) was used. The scale includes 12 items measuring two dimensions: exploration (e.g., “I spend time learning about my ethnic group's history and traditions”) and commitment (e.g., “I am proud of my ethnic group affiliation”). Responses were recorded on a 5-point Likert scale (1 = strongly disagree to 5 = strongly agree). The MEIM has demonstrated good reliability and validity across various cultural groups [30]. Social avoidance Social avoidance was measured using an adapted version of the Social Avoidance and Distress Scale (SADS; [74]). The original SADS uses a true/false response format. For the present study, the response format was modified to a 5-point Likert scale (1 = strongly disagree to 5 = strongly agree) to increase measurement sensitivity and facilitate latent profile analysis, which requires continuous indicators with sufficient variance. This modification follows the approach of several prior studies that have adapted social anxiety measures from dichotomous to Likert-type formats to improve measurement sensitivity and distributional properties [40, 61]. The adapted scale was pilot-tested with 45 African international students prior to the main study. Confirmatory factor analysis of the pilot data confirmed acceptable fit (CFI = 0.91, RMSEA = 0.07), and internal consistency was adequate (α = 0.84). To provide more rigorous validation, confirmatory factor analysis was also conducted in the main sample (N = 524). A single-factor model yielded borderline fit (χ2/df = 3.84, CFI = 0.86, TLI = 0.84, RMSEA = 0.074 [90% CI: 0.068, 0.080]). A two-factor model distinguishing avoidance items (8 items) from distress items (6 items) provided improved fit (χ2/df = 2.41, CFI = 0.93, TLI = 0.92, RMSEA = 0.052 [90% CI: 0.045, 0.059]). The two factors were highly correlated (r = 0.78), supporting the use of a composite score while acknowledging the two-dimensional structure of the adapted scale (see Supplementary Table S2). The main study used 14 items measuring individuals' avoidance tendencies in social situations (e.g., “I tend to avoid situations requiring socialization,” “I try not to talk to strangers”). Higher total scores indicate higher levels of social avoidance. Cultural intelligence The Cultural Intelligence Scale (CQS; [69]) was used. The CQS consists of 20 items measuring four dimensions: metacognitive CQ (4 items), cognitive CQ (6 items), motivational CQ (5 items), and behavioral CQ (5 items). Responses were recorded on a 7-point Likert scale (1 = strongly disagree to 7 = strongly agree). Notably, the CQS uses a 7-point scale whereas the other measures in this study use 5-point scales. Because LPA in Mplus estimates profiles based on raw means and variances within a maximum likelihood framework, standardization of indicators prior to analysis is not required and can in fact distort profile solutions by altering the variance structure [49]. Therefore, all indicators were entered in their original metric. Profile plots present standardized scores for visual comparability, but all model estimation used raw scores. To verify that the differing scale metrics did not disproportionately influence class formation, a sensitivity analysis was conducted in which all four indicators were z-standardized prior to LPA estimation. The three-profile solution remained optimal under the same selection criteria (BIC = 5475.45, entropy = 0.85, LMR-LRT p = 0.009; see Supplementary Tables S3 and S4). Profile sizes were highly comparable (34.5%, 20.8%, and 44.7%), and Cohen's kappa for classification agreement between the raw-metric and standardized-metric solutions was 0.91, indicating that the profile structure was robust to indicator scaling. Perceived discrimination The Perceived Discrimination Scale adapted for international students was used [35]. The scale includes 9 items measuring the frequency and severity of discrimination experiences in interpersonal, institutional, and public settings (e.g., “I have been treated unfairly because of my race or ethnicity”). Responses were recorded on a 5-point Likert scale (1 = never to 5 = very often). The scale has been validated among African international student populations [34]. Cross-cultural adaptation The revised Sociocultural Adaptation Scale (SCAS; [73]) was used. The scale includes 20 items measuring international students' ability to handle cross-cultural situations in daily life (e.g., “making Chinese friends,” “understanding Chinese people's values”). A 5-point Likert scale was used (1 = very difficult, 5 = very easy). The reliability and validity of SCAS have been widely validated in international student populations [75]. Covariates and auxiliary variables Based on previous research [4], demographic variables including gender (0 = female, 1 = male), age (continuous), time in China (continuous, in months), major type (dummy coded), and city of study (dummy coded) were measured. Chinese language proficiency (self-rated, 1 = not at all to 5 = very fluent) and social support (Multidimensional Scale of Perceived Social Support, MSPSS; [78]) were also measured. These variables served as covariates in profile comparisons and as correlates of profile membership. Common method bias control Since all variables were measured through self-report, the threat of common method bias (CMB) was addressed through both procedural and statistical strategies [37]. Procedural controls included anonymous responses, emphasis on no right or wrong answers, randomized item order, and varied scale formats. For statistical control, Harman's single-factor test was conducted: exploratory factor analysis (unrotated) showed that the first factor explained 26.8% of the variance, below the 40% critical value. Additionally, a confirmatory factor analysis comparing the measurement model with and without a common latent factor showed no significant improvement in model fit (Δχ2 = 18.4, df = 14, p > 0.05). We acknowledge that Harman's test has been criticized as insufficiently sensitive [3, 59], and the CFA-based test provides a more rigorous assessment. Together, the procedural and statistical evidence suggests that CMB was within an acceptable range, though it cannot be entirely ruled out. Qualitative research Interview sample selection The qualitative research adopted purposive sampling and maximum variation sampling strategies [52], selecting a subsample from quantitative research participants. Sample selection followed theoretical sampling principles based on preliminary LPA results, aiming to recruit participants from each identified latent profile. The selection ensured heterogeneity in profile membership, gender, country of origin, city of study, and time in China. The final interview sample comprised 18 participants (11 males and 7 females) from seven countries, including Nigeria (4), Kenya (3), Ghana (3), Ethiopia (3), Tanzania (2), South Africa (2), and Uganda (1). Duration of residence ranged from 8 months to 4 years. Six participants were selected from each of the three latent profiles. Participants were selected based on their most likely profile membership, defined as the profile to which they were assigned with the highest posterior probability. The average posterior probability for the 18 selected interviewees was 0.89 (range: 0.78 to 0.97), indicating adequate classification certainty for qualitative follow-up purposes. Participants were drawn from all four cities. Theoretical saturation was achieved by the 15th interview, with subsequent interviews confirming that no new themes emerged. Interview guide design and implementation The semi-structured interview guide contained four core modules: (1) identity recognition experience (e.g., “As an African international student in China, how do you view your identity?” “What experiences have influenced your understanding of your ethnic identity?”), (2) social interaction and avoidance (e.g., “Please describe your social interactions with Chinese students,” “Have you ever wanted to avoid certain social situations? Why?”), (3) cultural intelligence and discrimination experiences (e.g., “How have you learned to navigate cultural differences in China?” “Have you experienced any unfair treatment based on your background?”), and (4) cross-cultural adaptation process (e.g., “What is the biggest challenge you have encountered?” “How do you cope with these challenges?”). Interviews were conducted at times and locations convenient for participants, using English or languages familiar to interviewees, lasting 45 to 90 min each. All interviews were recorded and transcribed within 48 h. Transcripts were reviewed and confirmed by interviewees. Data analysis Quantitative data analysis Quantitative data analysis proceeded in four stages. In the first stage, preliminary analyses were conducted using SPSS 26.0, including descriptive statistics, bivariate correlations, and reliability analysis (Cronbach's α, composite reliability, and average variance extracted). In the second stage, LPA was conducted using Mplus 8.3 [25] with identity recognition, social avoidance, cultural intelligence, and perceived discrimination as indicator variables. Models specifying 1 through 5 profiles were estimated and compared using 500 random starts with 50 final-stage optimizations to ensure convergence to the global maximum. All models were estimated with within-profile variances freely estimated across profiles (i.e., class-varying variances), which allows profiles to differ not only in their means but also in the dispersion of indicators. This specification is consistent with recommendations for exploratory LPA applications where there is no strong a priori reason to constrain variances [56]. Model selection was guided by multiple fit indices: BIC and sample-size adjusted BIC (aBIC), the Lo-Mendell-Rubin adjusted likelihood ratio test (LMR-LRT), the Bootstrap Likelihood Ratio Test (BLRT), entropy, and minimum profile proportion. In cases of conflicting indices, BIC, BLRT, and interpretability were prioritized [53]. In the third stage, auxiliary analyses were conducted using the BCH method [23] to compare profiles on cross-cultural adaptation. The BCH method accounts for classification uncertainty by using weighted multiple-group comparisons, providing more accurate estimates than the classify-analyze approach [47]. Pairwise profile comparisons with Wald chi-square tests were conducted. In the fourth stage, multinomial logistic regression was conducted to examine correlates of profile membership. Demographic variables, Chinese language proficiency, and social support were entered as independent variables. Qualitative data analysis Qualitative data were analyzed using thematic analysis following Braun and Clarke's [19] six-step framework (see also [43]). The coding process was assisted by NVivo 12 [77]. Two researchers independently coded 20% of the transcripts (4 interviews), yielding an inter-coder agreement Kappa coefficient of 0.82. Differences were discussed and resolved before coding the remaining transcripts. The initial coding scheme was developed deductively from the four modules of the interview guide. During consensus discussions, two additional codes emerged inductively: “linguistic insecurity as avoidance motivation” and “identity compartmentalization.” These additions were applied prospectively to the remaining transcripts and did not require recoding of the four transcripts already analyzed, as both coders confirmed that the new codes captured content not previously assigned to existing categories. The analysis adopted a combined inductive and deductive strategy. Qualitative findings were mapped onto the quantitatively derived profiles. Ethical considerations All participants provided written informed consent to participate in this study. The study was conducted in accordance with the Declaration of Helsinki.This study was approved by the Ethics Committee of Wuxi University Of TechnologyAll participants were fully informed of the research purpose, procedures, potential risks and benefits, data confidentiality measures, and the right to withdraw at any time, and provided informed consent. For the interview component, it was emphasized that interviewees could refuse to answer any questions or terminate the interview without consequence. All data were anonymized, and pseudonyms were used in reporting. Results Descriptive statistics and correlation analysis Table 1 presents the means, standard deviations, and correlation coefficients among main variables. The mean identity recognition score was 3.62 (SD = 0.78), slightly above the scale midpoint, indicating moderately stable ethnic identity recognition. The social avoidance average score was 2.88 (SD = 0.86), at a moderate level. The cultural intelligence average score was 4.18 (SD = 0.92) on a 7-point scale, indicating moderate cultural intelligence. The perceived discrimination average score was 2.51 (SD = 0.94), below the scale midpoint but with notable variability. The cross-cultural adaptation average score was 3.34 (SD = 0.85).Table 1Descriptive statistics and correlation matrix of main variablesVariableMSD12345671. Identity Recognition3.620.78—2. Social Avoidance2.880.86-.39***—3. Cultural Intelligence4.180.92.42***-.35***—4. Perceived Discrimination2.510.94-.28***.38***-.22***—5. Cross-Cultural Adaptation3.340.85.48***-.46***.51***-.41***—6. Chinese Language Proficiency2.851.12.23***-.26***.36***-.19**.39***—7. Social Support4.561.08.34***-.31***.33***-.27***.43***.29***—N = 524*p < .05**p < .01***p < .001 Table 1Descriptive statistics and correlation matrix of main variablesVariableMSD12345671. Identity Recognition3.620.78—2. Social Avoidance2.880.86-.39***—3. Cultural Intelligence4.180.92.42***-.35***—4. Perceived Discrimination2.510.94-.28***.38***-.22***—5. Cross-Cultural Adaptation3.340.85.48***-.46***.51***-.41***—6. Chinese Language Proficiency2.851.12.23***-.26***.36***-.19**.39***—7. Social Support4.561.08.34***-.31***.33***-.27***.43***.29***—N = 524*p < .05**p < .01***p < .001 Descriptive statistics and correlation matrix of main variables N = 524 *p < .05 **p < .01 ***p < .001 Correlation analysis revealed that identity recognition was significantly positively correlated with cross-cultural adaptation (r = 0.48, p < 0.001) and cultural intelligence (r = 0.42, p < 0.001), and significantly negatively correlated with social avoidance (r = −0.39, p < 0.001) and perceived discrimination (r = −0.28, p < 0.001). Social avoidance was significantly negatively correlated with cross-cultural adaptation (r = −0.46, p < 0.001) and cultural intelligence (r = −0.35, p < 0.001), and significantly positively correlated with perceived discrimination (r = 0.38, p < 0.001). Cultural intelligence was significantly positively correlated with cross-cultural adaptation (r = 0.51, p < 0.001) but only weakly negatively correlated with perceived discrimination (r = −0.22, p < 0.001). This weaker-than-expected association suggests that cultural intelligence, while facilitating adaptation, may not substantially alter students' perceptions of discriminatory experiences. Perceived discrimination was significantly negatively correlated with cross-cultural adaptation (r = −0.41, p < 0.001). Notably, the correlation between identity recognition and social avoidance (r = −0.39) was moderate rather than strong, indicating that identity clarity alone may not be sufficient to prevent social avoidance without accompanying cultural competence resources. Among control variables, Chinese language proficiency (r = 0.39, p < 0.001) and time in China (r = 0.28, p < 0.001) were significantly positively correlated with cross-cultural adaptation. Reliability and validity testing Table 2 reports the reliability and validity indicators. All Cronbach's α coefficients ranged from 0.85 to 0.92, exceeding 0.70. Composite reliability (CR) ranged from 0.87 to 0.93, and average variance extracted (AVE) ranged from 0.48 to 0.56. The AVE for social avoidance (0.48) fell marginally below the recommended 0.50 threshold. Although CR exceeded 0.70 and discriminant validity was confirmed via the Fornell-Larcker criterion [26], this represents a limitation. The main-sample CFA (Supplementary Table S2) confirmed adequate fit for a two-factor structure, though the composite score was retained for LPA given the high inter-factor correlation (r = 0.78). Future research should further examine the factor structure of this adapted scale with independent samples. Discriminant validity was confirmed using the Fornell-Larcker criterion: each variable's square root of AVE exceeded its correlations with all other variables [1].Table 2Reliability and validity indicators of main scalesScaleItemsCronbach's αCRAVEIdentity Recognition (MEIM)12.87.89.50Social Avoidance (SADS-adapted)14.85.87.48Cultural Intelligence (CQS)20.91.93.53Perceived Discrimination9.86.88.51Cross-Cultural Adaptation (SCAS)20.92.93.56CR Composite Reliability, AVE Average Variance Extracted Table 2Reliability and validity indicators of main scalesScaleItemsCronbach's αCRAVEIdentity Recognition (MEIM)12.87.89.50Social Avoidance (SADS-adapted)14.85.87.48Cultural Intelligence (CQS)20.91.93.53Perceived Discrimination9.86.88.51Cross-Cultural Adaptation (SCAS)20.92.93.56CR Composite Reliability, AVE Average Variance Extracted Reliability and validity indicators of main scales CR Composite Reliability, AVE Average Variance Extracted Latent profile analysis Model selection LPA was conducted using identity recognition, social avoidance, cultural intelligence, and perceived discrimination as indicator variables. Models specifying 1 through 5 profiles were estimated. Table 3 presents the fit indices. The 3-profile solution was selected based on the following criteria: (a) BIC decreased substantially from the 2-profile to the 3-profile model but showed minimal improvement for the 4-profile model; (b) the LMR-LRT was significant for the 3-profile solution (p = 0.013) but nonsignificant for the 4-profile solution (p = 0.08); (c) entropy for the 3-profile model was 0.86, indicating clear profile separation; (d) the 4-profile model, while showing a significant BLRT, produced a smallest profile with only 7.8% of participants; and (e) the 3-profile solution was theoretically interpretable.Table 3Fit indices for latent profile models (1 to 5 profiles)ProfilesAICBICaBICLMR-LRT (p)BLRT (p)EntropyMin. %15842.165876.285849.73———100%25534.895590.475548.36.001<.001.8431.7%(n = 166)35312.745389.785332.11.013<.001.8621.4%(n = 112)45278.935377.435304.20.08<.001.837.8%(n = 41)55261.475381.435292.64.24<.001.804.2%(n = 22)Bold row indicates the selected model Table 3Fit indices for latent profile models (1 to 5 profiles)ProfilesAICBICaBICLMR-LRT (p)BLRT (p)EntropyMin. %15842.165876.285849.73———100%25534.895590.475548.36.001<.001.8431.7%(n = 166)35312.745389.785332.11.013<.001.8621.4%(n = 112)45278.935377.435304.20.08<.001.837.8%(n = 41)55261.475381.435292.64.24<.001.804.2%(n = 22)Bold row indicates the selected model Fit indices for latent profile models (1 to 5 profiles) Bold row indicates the selected model To ensure transparency, the four-profile solution is presented in Supplementary Table S1. In this solution, three of the four profiles closely replicated the three-profile structure. The fourth profile (n = 41, 7.8%) exhibited indicator means intermediate between the Cautious Navigators and Anxious Withdrawers profiles (identity recognition M = 3.21, social avoidance M = 3.47, cultural intelligence M = 3.51, perceived discrimination M = 3.06), representing a quantitative gradation rather than a qualitatively distinct configuration. The BIC improvement from the three-profile to the four-profile model was modest (a decrease of 12.35, compared to 200.69 from the two-profile to the three-profile model), and entropy declined from 0.86 to 0.83. We therefore retained the three-profile solution on the basis of parsimony, entropy, and substantive interpretability, while acknowledging that the BLRT remained significant for the four-profile model. Profile characteristics Table 4 presents the means of indicator variables across the three latent profiles, and Fig. 1 presents the standardized profile plot. The profiles were labeled based on their distinctive configurations:Table 4Means and standard deviations of indicator variables across latent profilesVariableProfile 1 (n = 186)Profile 2 (n = 112)Profile 3 (n = 226)FIdentity Recognition4.15 (0.54)2.82 (0.63)3.54 (0.59)198.74***Social Avoidance2.18 (0.61)3.89 (0.66)2.91 (0.72)224.31***Cultural Intelligence5.06 (0.73)3.28 (0.84)3.96 (0.78)186.52***Perceived Discrimination2.08 (0.71)3.34 (0.76)2.57 (0.83)108.46***Standard deviations in parentheses. Cultural intelligence measured on a 7-point scale; all other variables on a 5-point scale***p <.001Fig. 1Latent profile plot: standardized indicator means across three profiles Table 4Means and standard deviations of indicator variables across latent profilesVariableProfile 1 (n = 186)Profile 2 (n = 112)Profile 3 (n = 226)FIdentity Recognition4.15 (0.54)2.82 (0.63)3.54 (0.59)198.74***Social Avoidance2.18 (0.61)3.89 (0.66)2.91 (0.72)224.31***Cultural Intelligence5.06 (0.73)3.28 (0.84)3.96 (0.78)186.52***Perceived Discrimination2.08 (0.71)3.34 (0.76)2.57 (0.83)108.46***Standard deviations in parentheses. Cultural intelligence measured on a 7-point scale; all other variables on a 5-point scale***p <.001 Means and standard deviations of indicator variables across latent profiles Standard deviations in parentheses. Cultural intelligence measured on a 7-point scale; all other variables on a 5-point scale ***p <.001 Latent profile plot: standardized indicator means across three profiles Profile 1: Resilient Integrators (n = 186, 35.5%). This profile was characterized by high identity recognition (M = 4.15, SD = 0.54) and cultural intelligence (M = 5.06, SD = 0.73), low social avoidance (M = 2.18, SD = 0.61), and low perceived discrimination (M = 2.08, SD = 0.71). These students possessed strong ethnic identity alongside high cultural intelligence and actively engaged in cross-cultural interactions with minimal avoidance behavior.Profile 2: Anxious Withdrawers (n = 112, 21.4%). This profile was characterized by low identity recognition (M = 2.82, SD = 0.63) and cultural intelligence (M = 3.28, SD = 0.84), alongside high social avoidance (M = 3.89, SD = 0.66) and high perceived discrimination (M = 3.34, SD = 0.76). These students exhibited weak identity recognition combined with limited cultural intelligence, frequent social avoidance, and high perceived discrimination.Profile 3: Cautious Navigators (n = 226, 43.1%). This profile was characterized by moderate levels across all four indicators: identity recognition (M = 3.54, SD = 0.59), social avoidance (M = 2.91, SD = 0.72), cultural intelligence (M = 3.96, SD = 0.78), and perceived discrimination (M = 2.57, SD = 0.83). Importantly, this profile exhibited the largest within-profile standard deviations on perceived discrimination and social avoidance, suggesting greater heterogeneity than the other two profiles. Profile 1: Resilient Integrators (n = 186, 35.5%). This profile was characterized by high identity recognition (M = 4.15, SD = 0.54) and cultural intelligence (M = 5.06, SD = 0.73), low social avoidance (M = 2.18, SD = 0.61), and low perceived discrimination (M = 2.08, SD = 0.71). These students possessed strong ethnic identity alongside high cultural intelligence and actively engaged in cross-cultural interactions with minimal avoidance behavior. Profile 1: Resilient Integrators (n = 186, 35.5%). This profile was characterized by high identity recognition (M = 4.15, SD = 0.54) and cultural intelligence (M = 5.06, SD = 0.73), low social avoidance (M = 2.18, SD = 0.61), and low perceived discrimination (M = 2.08, SD = 0.71). These students possessed strong ethnic identity alongside high cultural intelligence and actively engaged in cross-cultural interactions with minimal avoidance behavior. Profile 2: Anxious Withdrawers (n = 112, 21.4%). This profile was characterized by low identity recognition (M = 2.82, SD = 0.63) and cultural intelligence (M = 3.28, SD = 0.84), alongside high social avoidance (M = 3.89, SD = 0.66) and high perceived discrimination (M = 3.34, SD = 0.76). These students exhibited weak identity recognition combined with limited cultural intelligence, frequent social avoidance, and high perceived discrimination. Profile 2: Anxious Withdrawers (n = 112, 21.4%). This profile was characterized by low identity recognition (M = 2.82, SD = 0.63) and cultural intelligence (M = 3.28, SD = 0.84), alongside high social avoidance (M = 3.89, SD = 0.66) and high perceived discrimination (M = 3.34, SD = 0.76). These students exhibited weak identity recognition combined with limited cultural intelligence, frequent social avoidance, and high perceived discrimination. Profile 3: Cautious Navigators (n = 226, 43.1%). This profile was characterized by moderate levels across all four indicators: identity recognition (M = 3.54, SD = 0.59), social avoidance (M = 2.91, SD = 0.72), cultural intelligence (M = 3.96, SD = 0.78), and perceived discrimination (M = 2.57, SD = 0.83). Importantly, this profile exhibited the largest within-profile standard deviations on perceived discrimination and social avoidance, suggesting greater heterogeneity than the other two profiles. Profile 3: Cautious Navigators (n = 226, 43.1%). This profile was characterized by moderate levels across all four indicators: identity recognition (M = 3.54, SD = 0.59), social avoidance (M = 2.91, SD = 0.72), cultural intelligence (M = 3.96, SD = 0.78), and perceived discrimination (M = 2.57, SD = 0.83). Importantly, this profile exhibited the largest within-profile standard deviations on perceived discrimination and social avoidance, suggesting greater heterogeneity than the other two profiles. Profile differences in cross-cultural adaptation The BCH method was used to compare cross-cultural adaptation across profiles. Table 5 presents the results. Overall, the three profiles differed significantly (χ2 = 156.82, df = 2, p < 0.001). Profile 1 (Resilient Integrators, M = 3.98, SE = 0.06) reported significantly higher adaptation than both Profile 3 (Cautious Navigators, M = 3.31, SE = 0.05; χ2 = 68.34, p < 0.001) and Profile 2 (Anxious Withdrawers, M = 2.58, SE = 0.08; χ2 = 142.67, p < 0.001). Profile 3 also reported significantly higher adaptation than Profile 2 (χ2 = 51.23, p < 0.001). The gradient was clear: Resilient Integrators > Cautious Navigators > Anxious Withdrawers. Notably, the effect size for the Profile 1 versus Profile 3 comparison (Cohen's d = 0.86) was substantially smaller than for Profile 1 versus Profile 2 (Cohen's d = 1.69), suggesting that the difference between the top two profiles, while significant, was more modest than the gap separating both from the vulnerable group.Table 5BCH analysis: profile differences in cross-cultural adaptationComparisonM (SE)M (SE)Wald χ2dfpOverall——156.822<.001Profile 1 vs. Profile 23.98 (0.06)2.58 (0.08)142.671<.001Profile 1 vs. Profile 33.98 (0.06)3.31 (0.05)68.341<.001Profile 2 vs. Profile 32.58 (0.08)3.31 (0.05)51.231<.001Profile 1 Resilient Integrators, Profile 2 Anxious Withdrawers, Profile 3 Cautious Navigators Table 5BCH analysis: profile differences in cross-cultural adaptationComparisonM (SE)M (SE)Wald χ2dfpOverall——156.822<.001Profile 1 vs. Profile 23.98 (0.06)2.58 (0.08)142.671<.001Profile 1 vs. Profile 33.98 (0.06)3.31 (0.05)68.341<.001Profile 2 vs. Profile 32.58 (0.08)3.31 (0.05)51.231<.001Profile 1 Resilient Integrators, Profile 2 Anxious Withdrawers, Profile 3 Cautious Navigators BCH analysis: profile differences in cross-cultural adaptation Profile 1 Resilient Integrators, Profile 2 Anxious Withdrawers, Profile 3 Cautious Navigators Correlates of profile membership Multinomial logistic regression was conducted with Profile 3 (Cautious Navigators) as the reference group. Table 6 presents the results. Compared to Cautious Navigators, membership in the Resilient Integrators profile was significantly associated with higher Chinese language proficiency (OR = 1.74, 95% CI [1.34, 2.26], p < 0.001) and higher social support (OR = 1.58, 95% CI [1.22, 2.05], p < 0.001). Time in China approached significance (OR = 1.02, 95% CI [1.00, 1.04], p = 0.09). Gender and age were not significantly associated with profile membership. City of study was also examined as a covariate; students in Guangzhou showed a marginally higher likelihood of Resilient Integrators membership relative to Wuxi (OR = 1.38, 95% CI [0.98, 1.94], p = 0.07), possibly reflecting the larger and more established African community in Guangzhou, though this effect did not reach significance.Table 6Multinomial logistic regression: correlates of profile membership (Reference: Profile 3)PredictorP1 vs. 3 OR95% CIpP2 vs. 3 OR95% CIpGender1.18[0.78, 1.79].430.91[0.56, 1.48].71Age1.03[0.94, 1.13].520.97[0.87, 1.08].58Time in China1.02[1.00, 1.04].090.97[0.95, 0.99].01Chinese Language1.74[1.34, 2.26]<.0010.54[0.40, 0.73]<.001Social Support1.58[1.22, 2.05]<.0010.59[0.43, 0.81].001City of study was included as a covariate (not shown for brevity). Time in China was measured in months; the OR of 1.02 indicates a 2% increase in the odds of Resilient Integrators membership per additional monthProfile 1 = Resilient Integrators; Profile 2 = Anxious Withdrawers; Profile 3 = Cautious NavigatorsOR Odds Ratio, CI Confidence Interval Table 6Multinomial logistic regression: correlates of profile membership (Reference: Profile 3)PredictorP1 vs. 3 OR95% CIpP2 vs. 3 OR95% CIpGender1.18[0.78, 1.79].430.91[0.56, 1.48].71Age1.03[0.94, 1.13].520.97[0.87, 1.08].58Time in China1.02[1.00, 1.04].090.97[0.95, 0.99].01Chinese Language1.74[1.34, 2.26]<.0010.54[0.40, 0.73]<.001Social Support1.58[1.22, 2.05]<.0010.59[0.43, 0.81].001City of study was included as a covariate (not shown for brevity). Time in China was measured in months; the OR of 1.02 indicates a 2% increase in the odds of Resilient Integrators membership per additional monthProfile 1 = Resilient Integrators; Profile 2 = Anxious Withdrawers; Profile 3 = Cautious NavigatorsOR Odds Ratio, CI Confidence Interval Multinomial logistic regression: correlates of profile membership (Reference: Profile 3) City of study was included as a covariate (not shown for brevity). Time in China was measured in months; the OR of 1.02 indicates a 2% increase in the odds of Resilient Integrators membership per additional month Profile 1 = Resilient Integrators; Profile 2 = Anxious Withdrawers; Profile 3 = Cautious Navigators OR Odds Ratio, CI Confidence Interval Compared to Cautious Navigators, membership in the Anxious Withdrawers profile was significantly associated with lower Chinese language proficiency (OR = 0.54, 95% CI [0.40, 0.73], p < 0.001), lower social support (OR = 0.59, 95% CI [0.43, 0.81], p = 0.001), and shorter time in China (OR = 0.97, 95% CI [0.95, 0.99], p = 0.01). The overall model was significant (χ2 = 124.56, df = 14, p < 0.001, Nagelkerke R2 = 0.26). Qualitative findings Thematic analysis of interviews with 18 participants revealed three core themes that contextualized the quantitative findings. While qualitative results largely corroborated the profile structure, they also revealed important nuances and within-profile variability that the quantitative data alone could not capture. Theme 1: divergent identity configurations and their experiential basis Interviews revealed that the three profiles differed in how identity recognition interacted with cultural intelligence and discrimination experiences. Participants corresponding to the Resilient Integrators profile described identity recognition as a dynamic resource deployed in cross-cultural encounters. A Nigerian participant in Beijing (P03) explained: “I know who I am. I am Nigerian, I am African, and I am also an international student here. When I understand Chinese culture better, it does not weaken my identity; it makes me richer.” This narrative illustrates how strong identity recognition combined with high cultural intelligence created a synergistic resource for engagement. However, not all Resilient Integrators described a uniformly smooth trajectory. A Ghanaian participant in Guangzhou (P04) shared: “When I first came, I actually withdrew a lot. I was confused. But after about six months, I joined a language exchange group and things changed. Once I started understanding the culture, my confidence came back.” This account suggests that some Resilient Integrators may have transitioned from a more avoidant pattern over time, a dynamic that the cross-sectional quantitative data could not capture. Participants corresponding to the Anxious Withdrawers profile described identity as a source of vulnerability compounded by discrimination experiences. A Kenyan participant in Wuhan (P09) shared: “Sometimes I feel confused about who I am here. I am proud to be Kenyan, but when people stare at me on the bus or move away from me, I start wondering if there is something wrong with me.” This account illustrates the co-occurrence of low ethnic identity commitment, limited cultural intelligence, and high perceived discrimination in a pattern of compounded vulnerability. Participants corresponding to the Cautious Navigators profile described a more ambivalent experience. A Ghanaian participant in Wuxi (P16) reflected: “I am generally comfortable with who I am, but I often feel unsure about how to act in Chinese social situations. I do not avoid people, but I am careful.” However, some Cautious Navigators described experiences that partially overlapped with the Anxious Withdrawers pattern. An Ethiopian participant in Beijing (P14) noted: “Most days I manage fine, but there are periods, maybe after a bad experience, where I pull back completely for a week or two. Then I push myself to go out again.” This within-profile variability underscores that the Cautious Navigators category may encompass a broader range of experiences than the moderate quantitative scores suggest. Further analysis of the interview data revealed that ethnic identity, as captured by the MEIM, operated through qualitatively different functions across participants, even among those with similarly high quantitative scores. Three distinct functions were identified. For some Resilient Integrators, ethnic identity operated primarily as a source of pride and positive distinctiveness. A Nigerian participant in Beijing (P03) elaborated: “Being Nigerian gives me confidence. We are resilient people, we are entrepreneurial, we adapt. I carry that with me into every interaction here.” This function aligns with social identity theory's emphasis on positive group evaluation as a psychological resource. For others, ethnic identity functioned as a form of active resistance to perceived marginalization. A Tanzanian participant in Guangzhou (P07) described: “Sometimes when I feel looked down upon, I remind myself of my education, my culture, my language. I refuse to accept that I am less than anyone here. My identity is my shield.” This function resonates with the rejection-identification model, in which discrimination strengthens rather than weakens group identification. A third function, observed primarily among Cautious Navigators, was pragmatic compartmentalization. A South African participant in Guangzhou (P15) explained: “I am proudly South African, but I have learned to put that aside when I need to. In class, I try to just be a student, not an African student. I switch between identities depending on the situation.” This selective deployment of ethnic identity was not captured by the MEIM's continuous measure but may represent an important adaptive strategy in high cultural-distance contexts. These varied identity functions suggest that the same quantitative score on the MEIM can reflect qualitatively different psychological processes with different implications for cross-cultural engagement. This finding complements the quantitative results by revealing that ethnic identity's contribution to adaptation configurations is more nuanced than a single continuous dimension can capture. Theme 2: the interplay of social avoidance, discrimination, and cultural competence A second theme concerned the complex interplay between social avoidance, perceived discrimination, and cultural intelligence. For Resilient Integrators, high cultural intelligence appeared to serve as a buffer against avoidance. A Tanzanian participant in Guangzhou (P07) described: “Of course there are sometimes uncomfortable moments. Once a shop owner followed me around the store. But I have learned that this is not always personal. Sometimes it is just curiosity or unfamiliarity. Understanding the cultural context helps me not take it too personally and not withdraw.” For Anxious Withdrawers, the absence of this interpretive buffer meant that discrimination experiences were more likely to trigger avoidance. An Ethiopian participant in Wuhan (P12) shared: “Every time something negative happens, I think this is because I am African. I do not know enough about Chinese culture to understand if this is normal behavior or actual discrimination. So I just avoid going out unless I have to.” Importantly, the relationship between discrimination and avoidance was not uniform across cities. Several participants in Guangzhou noted that the larger African community provided a social buffer. A South African participant in Guangzhou (P06) observed: “In Guangzhou, there are many Africans, so you do not stand out as much. I think students in smaller cities where they are more visible probably have a harder time.” This geographic variation was partially reflected in the quantitative data, where Guangzhou students showed marginally higher Resilient Integrators membership, though the effect did not reach significance. Cautious Navigators exhibited a selective avoidance pattern. A Ugandan participant in Beijing (P18) explained: “I do not avoid all social situations, but I am selective. I am comfortable in class or in structured activities. But unstructured social events like parties make me nervous because I am not sure about the social rules.” Beyond these profile-level differences, the interview data revealed that social avoidance was not a unitary phenomenon but was driven by distinct motivations that varied both across and within profiles. Four distinguishable motivations emerged from the data. The first was anxiety-driven avoidance, in which participants described a generalized apprehension about social encounters. A Kenyan participant in Wuhan (P09) explained: “It is not that something specific happened. I just feel nervous around people I do not know, especially when I cannot express myself well. My heart races and I want to leave.” This pattern was most common among Anxious Withdrawers but was also described by two Cautious Navigators during periods of heightened stress. The second was discrimination-triggered withdrawal, in which avoidance was a direct response to specific negative experiences. An Ethiopian participant in Wuhan (P12) described: “After a taxi driver refused to pick me up twice in one week, I stopped trying to go out on my own. I only go places now if a Chinese classmate comes with me.” This motivation was concentrated among Anxious Withdrawers but was also reported by Cautious Navigators following particularly salient negative encounters. The third was linguistic insecurity, in which limited Chinese proficiency made social interaction effortful and embarrassing. A Ghanaian participant in Wuxi (P16) shared: “I avoid the canteen during busy hours because I cannot read the menu fast enough and people behind me get impatient. It is not fear exactly; it is more like shame.” This motivation appeared across all three profiles, including among some Resilient Integrators in their earlier months in China. The fourth was strategic selectivity, in which participants deliberately limited their social exposure to conserve energy or prioritize academic demands. A Ugandan participant in Beijing (P18) noted: “I have limited social energy. I choose to spend it on situations where I know I will learn something or build a useful relationship. Random socializing feels like a waste when I have so much coursework.” This motivation was most characteristic of Cautious Navigators and was distinct from the other three in that it did not reflect distress or negative affect. These four motivations have different theoretical and practical implications. Anxiety-driven avoidance may respond to cognitive-behavioral interventions, discrimination-triggered withdrawal points to the need for environmental and institutional change, linguistic insecurity highlights the role of language training, and strategic selectivity may represent an adaptive response requiring no intervention. The co-existence of multiple motivations within a single quantitative profile, particularly the Cautious Navigators, helps explain the within-profile heterogeneity observed in the quantitative data. Theme 3: pathways to adaptation across profiles The third theme illuminated adaptation pathways characterizing each profile. Resilient Integrators described a mutually reinforcing pattern of engagement and cultural learning. A Nigerian participant in Wuxi (P02) shared: “The more I interact with Chinese classmates, the more I understand the culture, and the more confident I feel. Each positive experience gives me energy to reach out more.” Anxious Withdrawers described the opposite pattern. A Kenyan participant in Beijing (P10) reflected: “When I first arrived, I tried to make Chinese friends. But after some bad experiences, I started staying more with my Kenyan friends. Now, two years later, my Chinese is still poor, I do not understand the culture, and I feel even more isolated than before. I know avoiding was not the answer, but I did not know what else to do.” This account illustrates how multiple risk factors co-occurred and were experienced by the participant as mutually reinforcing. Cautious Navigators described a gradual and uneven trajectory. A Ghanaian participant in Wuhan (P17) noted: “My adaptation has been slow but steady. Some weeks I feel like I am making progress, other weeks I feel stuck. I wish I had more tools to understand the culture better.” Notably, two Cautious Navigator participants (P14 and P15) described periods of stagnation lasting several months, suggesting that the moderate resource profile may produce plateau effects. A South African participant in Guangzhou (P15) observed: “I feel like I have been at the same level of comfort for almost a year now. Not bad, not great. I wonder if this is as good as it gets without really pushing myself.” Discussion This study adopted a person-centered approach to identify distinct latent profiles among African international students in China based on their configurations of identity recognition, social avoidance, cultural intelligence, and perceived discrimination. Three profiles emerged that differed significantly in cross-cultural adaptation outcomes. The mixed-methods design enriched these findings by revealing not only the experiences underlying profile membership but also important within-profile heterogeneity and geographic variation. Theoretical interpretation of core findings The three-profile solution contributes to cross-cultural adaptation theory in several ways. Before discussing each profile, it is important to note that because perceived discrimination was included as a profile indicator, the resulting profiles represent configurations of both psychological characteristics and experienced environmental conditions rather than purely individual typologies. This distinction has implications for interpretation and for intervention design, as discussed below. The Resilient Integrators profile illustrates a synergistic pattern: strong ethnic identity commitment, high cultural intelligence, low avoidance, and low perceived discrimination co-occurred in a pattern consistent with mutually reinforcing processes of engagement and adaptation, though the cross-sectional design does not permit confirmation of the cyclical directionality. This finding is consistent with social identity theory, which suggests that a well-explored and committed ethnic identity may provide a psychological foundation for cross-group engagement [31], and extends it by indicating that cultural intelligence may serve as a complementary resource that facilitates the translation of ethnic identity commitment into effective cross-cultural interaction. The Anxious Withdrawers profile illustrates a co-occurring vulnerability pattern. Low ethnic identity commitment, limited cultural intelligence, high social avoidance, and high perceived discrimination were observed together in a configuration suggestive of mutually reinforcing processes, though the cross-sectional design precludes establishing whether these factors amplify one another over time. This extends the stress-coping literature [38] by demonstrating that avoidance behavior in cross-cultural contexts is embedded in a constellation of co-occurring risk factors rather than being a response to a single stressor. The rejection-identification model [18] offers a useful lens for understanding this profile. According to this model, perceived discrimination can either strengthen in-group identification as a protective response or, when identity resources are limited, contribute to identity confusion and withdrawal. The co-occurrence of high perceived discrimination and low ethnic identity commitment in the Anxious Withdrawers profile is consistent with the latter pathway. Furthermore, Borinca et al. [16] have shown that negative contact experiences can intensify intergroup anxiety and reduce openness to subsequent cross-group interaction, suggesting that the social avoidance observed in this profile may be partly maintained by affective responses to prior negative encounters rather than solely by dispositional tendencies. The Cautious Navigators profile, the largest subgroup at 43.1%, is perhaps the most theoretically informative. These students exhibited moderate levels across all indicators. Their qualitative narratives revealed selective engagement, approaching structured situations while avoiding unstructured ones. However, the qualitative data also revealed that this profile encompassed a broader range of experiences than the quantitative means suggest, with some members describing periodic withdrawal resembling the Anxious Withdrawers pattern. This within-profile heterogeneity suggests that the Cautious Navigators category may encompass students whose experiences vary along a continuum, with some resembling the Resilient Integrators pattern and others resembling the Anxious Withdrawers pattern. Whether these differences reflect developmental progression or stable individual variation cannot be determined from cross-sectional data. More broadly, the within-profile heterogeneity observed in this group raises the question of whether the three-profile solution captures discrete qualitative types or regions of concentration along a severity continuum. We acknowledge that LPA identifies areas of relative density in a multivariate distribution and does not, by itself, prove the existence of categorically distinct types [9]. The present profiles are therefore best interpreted as heuristically useful subgroup descriptions that identify recognizable configurations, rather than as evidence for taxonomic boundaries. Future research using taxometric methods or longitudinal designs would help clarify whether these configurations represent stable categories or points along a continuum. One unexpected quantitative finding deserves attention. The correlation between cultural intelligence and perceived discrimination (r = −0.22) was weaker than anticipated, given that higher cultural intelligence theoretically should reduce threat perception and therefore perceived discrimination. This weak association may reflect the nature of the discrimination African students face in China. Discrimination based on visible racial characteristics may be less amenable to reappraisal through cultural knowledge than discrimination based on cultural misunderstandings or behavioral differences. Qualitative data supported this interpretation: even highly culturally intelligent participants (e.g., P07) described incidents that were clearly racially motivated and not easily reinterpreted through cultural knowledge. This finding underscores the importance of distinguishing between discrimination that arises from cultural misunderstanding, which may be partially mitigated through cultural competence, and discrimination rooted in racial categorization, which reflects structural and attitudinal conditions less amenable to individual-level coping. Borinca, Guerra, and Uka [15] similarly observed that responses to intergroup exclusion are shaped by identity-related processes and belonging needs in ways that cannot be fully accounted for by individual competence alone. Furthermore, Borinca, Tropp, and Ofosu [17] demonstrated that meta-perceptions of how the outgroup views one's ingroup can shape willingness to engage in cross-group interaction, suggesting that African students' perceptions of how Chinese host nationals view their group may influence withdrawal or openness independently of individual-level cultural competence. Theoretical contributions Our first theoretical contribution lies in demonstrating the value of person-centered approaches for cross-cultural adaptation research. Previous studies have predominantly used variable-centered methods to examine average relationships between adaptation variables [11, 44, 71]. Our results reveal that the same variables can combine in qualitatively different configurations that represent distinct adaptation patterns with different underlying dynamics. Our second contribution is the integration of cultural intelligence and perceived discrimination into cross-cultural adaptation profiles. While cultural intelligence has been studied extensively as an individual predictor [63], it has not been examined as part of a profile configuration alongside identity recognition and social avoidance. Our findings suggest that cultural intelligence operates as a crucial moderating resource within profiles: present alongside strong identity (Profile 1), it enables confident engagement; absent alongside weak identity (Profile 2), it removes a critical buffer against avoidance-inducing experiences. Importantly, because perceived discrimination is included as an indicator, the profiles capture how individual resources interact with experienced environmental conditions. This approach aligns with ecological perspectives on adaptation [14] and avoids the assumption that adaptation is determined solely by individual-level attributes. Third, by recruiting participants from four geographically diverse Chinese cities, we reduce the likelihood that the identified profile structure is an artifact of a single institutional context. However, in the absence of formal cross-site invariance testing, structural generalizability cannot be confirmed and remains an important direction for future research. Practical implications The following practical implications should be interpreted as preliminary and conditional. The profile solution has not been cross-validated, no intervention was tested in this study, and the stability of profile membership over time is unknown. With these caveats in mind, the identification of three profiles suggests several directions for future applied research and institutional consideration. For students resembling Anxious Withdrawers, multi-component interventions addressing identity support, cultural intelligence training, structured social exposure, and discrimination coping strategies are indicated. The association between low Chinese language proficiency, low social support, and membership in this profile suggests that, pending replication, these variables may serve as potential indicators for identifying students who could benefit from enhanced support. However, whether such identification can be reliably implemented in routine orientation procedures remains to be tested. For students resembling Cautious Navigators, cultural intelligence training and facilitated low-stakes social opportunities may be most effective. Programs such as structured cultural exchange activities, buddy systems with Chinese students, and guided reflection workshops could help these students develop the competence to engage more confidently in unstructured contexts. Given the within-profile heterogeneity revealed in the qualitative data, periodic reassessment of these students' adaptation status may be warranted to identify those at risk of declining toward the Anxious Withdrawers pattern. For Resilient Integrators, the focus should shift to maintenance and peer leadership. The association between Chinese language proficiency and profile membership is consistent with the view that language training may serve as a foundational support resource, though causal direction cannot be established from these cross-sectional data. An important caveat applies to all of the above recommendations. Because the profiles incorporate perceived discrimination as an indicator, they partly reflect environmental conditions rather than solely individual characteristics. Accordingly, interventions should operate at multiple levels. Individual-level support, such as identity exploration activities, cultural intelligence training, and structured social exposure, addresses the psychological and behavioral components of the profiles. However, reducing the discrimination component requires institutional and systemic interventions, including anti-discrimination policies, diversity and inclusion training for host-national students and staff, and proactive efforts to foster welcoming campus climates. Focusing exclusively on individual-level change without addressing the environmental conditions that shape perceived discrimination would risk placing the burden of adaptation entirely on the students themselves. It should be noted that there is partial conceptual overlap between the profile indicators and the distal outcome. Cultural intelligence, social avoidance, and perceived discrimination are not fully independent of cross-cultural adaptation in a theoretical sense. The finding that profiles characterized by more favorable indicator configurations also reported higher adaptation is therefore partly built into the analytic design and should be interpreted with this constraint in mind. The more informative finding may be the relative magnitude of the differences: the smaller-than-expected gap between Resilient Integrators and Cautious Navigators suggests that moderate resources can yield reasonable adaptation in structured contexts. Limitations and future directions Several limitations should be noted. First, the cross-sectional design precludes examination of how profiles develop and change over time. The qualitative finding that some Resilient Integrators described earlier periods of withdrawal (e.g., P04) suggests that profile membership may be dynamic. Longitudinal research tracking profile transitions would clarify this. Second, while the multi-site design improved geographic representativeness, generalizability to other host countries and cultural contexts remains to be tested. Third, measurement relied on self-report instruments. While procedural and statistical controls were implemented, future studies should incorporate behavioral measures (e.g., social network analysis, experience sampling) and informant reports. Fourth, the adapted SADS scale, while pilot-tested, would benefit from further validation research with larger independent samples. Fifth, the MEIM captures ethnic identity exploration and commitment but does not assess bicultural identity integration, host-culture identification, or acculturative identity negotiation. Future research incorporating these additional identity dimensions would provide a more comprehensive understanding of how identity processes contribute to adaptation configurations. Sixth, the present study focused on four indicator variables. Future research could explore additional indicators such as psychological resilience, acculturative stress, or host country attitudes. Finally, the practical effectiveness of profile-based interventions remains to be tested through experimental or quasi-experimental designs. Conclusion This mixed-methods study adopted a person-centered approach to identify three distinct latent profiles among African international students in China: Resilient Integrators, Cautious Navigators, and Anxious Withdrawers. These profiles were characterized by qualitatively different configurations of identity recognition, social avoidance, cultural intelligence, and perceived discrimination, and differed significantly in cross-cultural adaptation outcomes. The findings indicate that international students do not form a homogeneous group; rather, they exhibit distinct patterns associated with different levels of cross-cultural adaptation. Chinese language proficiency and social support were significantly associated with profile membership, suggesting potential areas for further investigation in intervention research. Qualitative interviews revealed not only the experiential depth underlying these profiles but also important within-profile variability and geographic nuances that point to avenues for future research. Supplementary Information Supplementary Material 1. Supplementary Material 2. Supplementary Material 1. Supplementary Material 2.