My earliest memories of divination are rooted in the noisy, overlapping voices of home. In a village near Meizhou, Guangdong, Bazi readings are communal and intergenerational. I once accompanied my mother and relatives to see a local shenpo, whose reputation was part of our family’s oral tradition. The journey—thirty minutes up muddy mountain roads—felt like a return to something older than memory. Her home was modest, but one room stood apart: an altar thick with incense, talismans, and statues of Guanyin. The air carried joss smoke and damp soil—touch, smell, atmosphere. We sat together—cousins, aunts, uncles—sharing questions about marriage, health, and ancestral graves.
When AI Reads Your Fate: Bazi Divination and Algorithmic Life on the Cece App
What happens when an app claims to read your destiny?
Field Site: Cece App
In today’s China, a new wave of digital platforms is reinventing ancient practices like Bazi (八字) divination—a system that interprets your fate based on the time and date of your birth. Among them, the Cece (测测) app stands out. With over 36 million users, it combines traditional Chinese cosmology with sleek UX design and artificial intelligence.
But Cece is more than just a fortune-telling app. It positions itself as an "emotional support platform," blending AI-driven diagnostics with psychology quizzes, personality assessments, and daily life advice. Beneath its polished surface lies a deeper story—one about how prediction, governance, and personal meaning are being reshaped in a datafied world.
Drawing on my ethnographic research, this project explores how users navigate Cece’s algorithmic readings: some accept them, many critique them, and others co-create meaning together in vibrant comment sections. What emerges is not just a story of ancient traditions meeting new technology, but of how people reclaim agency in a world increasingly governed by opaque systems.
This piece sits at the intersection of anthropology, technology, and design. It asks:
What happens when AI replaces the oracle? And how do people push back—not by rejecting it, but by reinterpreting it together?
Commodification, Algorithm, and Communication
In China’s digital landscape, divination practices have evolved to embrace AI-driven platforms. The Cece (测测) app exemplifies this shift, navigating China’s regulatory environment by positioning itself not as a divination service but as “an online emotional companionship platform, leveraging AI technology, combining with human input to provide users with scientifically-based, warm, and engaging emotional counseling and support services.” (From Cece’s official website) This strategic framing allows Cece to thrive despite official attitudes toward “feudal superstition,”claiming instead to help users “achieve personal growth and emotional well-being.”
Upon opening the CeCe app, the user is greeted by a neatly designed, color-coded dashboard—a visual system that resembles a weather forecast. This interface breaks fate into four discrete categories: love, wealth, career, and study, each represented by a vertical bar with a numerical score. A central “composite score” summarizes overall wellbeing, while a short textual diagnosis advises users on how to “improve efficiency” by adjusting their mood or mental state. Comparing this figure with the one I showed in the previous section (see Fig.2)—the more traditional chart, here is a different pattern: There are no fixed numbers, no central “composite score.” Instead, the diviner inscribes four pillars—the year, month, day, and hour—each composed of heavenly stems and earthly branches. These components are relational. This feature stands out visually is the presence of arrows and loops that weave across the board, connecting different elements to one another. The diviner draws lines not to calculate, but to narrate—to show how one element may generate, overcome, or destabilize another within a broader pattern of cyclical time.
The app’s default page showcases a collection of psychology-oriented features: MBTI personality assessments, astrological profiles, compatibility tests, an interactive “Psychological Sandbox,” confidential sharing spaces, and dedicated “Psychological Rooms” for themed discussions. The section for Bazi divination is hidden in second page.
Navigating to the “生辰” (Bazi Birth Data) section, users encounter a summary of their Bazi interpretation. At the center is a cyclical diagram mapping out the relationships of 生 (generation) and 克 (control) between the Five Elements, a visual representation of the internal dynamics shaping one’s fate. Yet beneath is a small disclaimer lingers in understated font: This current content is free, only for entertainment. This is not professional analysis, does not represent value judgment, and is not an instruction for real-life decisions. On one hand, it functions as a standard disclaimer, distancing the platform from responsibility for users’ life choices. Yet on the other hand, it also reflects a deeper epistemic shift in how Bazi knowledge is framed. Here, “free” is not simply an economic category—it becomes synonymous with unprofessionalism, marking a division between algorithmically generated interpretations and the “real,” expert-guided Bazi reading, which exists behind a paywall. Knowledge is stratified, its perceived authenticity and value determined by the user’s willingness to pay.
Moreover, the interface is well designed to redirect users from free services toward paid options. As shown in Figure 6, the app generates enticing questions like “when will I get rich (发财)” for users. Clicking this redirects the user to a payment gateway, where access to an answer requires topping up credits. Here, expertise itself is graded and priced: users can choose from different tiers of consultants, with higher-level “masters” commanding higher fees. The platform thus orchestrates a user journey from algorithmic play to financial transaction, where speculative desire
is skillfully converted into microeconomic behavior.
Scrolling down, users are presented with a detailed AI-generated interpretation of their Bazi characteristics. Each element of their birth chart comes with a machine-processed explanation, stripped of the human touch of a traditional Bazi diviner. However, below these AI-generated readings is a commentary section where users can engage with one another—sharing experiences, asking questions, debating interpretations.
While the divination process itself has been automated, or directed into paid options, the knowledge-making process remains social. There is a commentary area for users to discuss in which they collectively negotiate its meaning, reinterpret its relevance, and re-embed it within their own lived realities. Cece opens up a more participatory, decentralized mode of knowledge production, where users themselves contribute to the shaping and validation of fate.
Cece operates at the intersection of algorithmic standardization, commercial stratification, and interactive meaning-making. While its design steers users toward standardized or paid interpretations, many instead engage in communal negotiation.
Shared Experience and Algorithm Logic
The AI-generated analysis complicates the question of what qualifies as divinatory knowledge. Traditionally, Bazi divination relies heavily on relational interpretations and contextual nuances provided by human diviners. In contrast, algorithmic divination operates through computational logic—essentially, algorithms as sets of instructions guiding calculations and operations. Within computer science, these algorithms represent programmed steps instructing computers on how to learn, process data, and generate outputs autonomously. Thus, AI-generated divination shifts Bazi readings from relational and interpretive practices toward standardized computational procedures, fundamentally reshaping how divinatory knowledge is defined, produced, and authenticated.
However, indeed, one way to reinterpret Bazi divination is by drawing parallels with statistical probability and algorithmic reasoning. Viewed through this lens, Bazi functions as an ancient form of data analysis—a system meticulously observed, recorded, and refined across generations. The patterns derived from the Five Elements (五行) and the Yin-Yang framework resemble statistical trends and probabilities. Categorizing individuals based on their elemental compositions from birth data mirrors the logic of contemporary predictive algorithms, where input variables produce forecasts of future trajectories. Consequently, Bazi’s cosmological structure can be understood as an early computational model (Chu, 2019, 214)—transforming complex life patterns into quantifiable and actionable insights about individual destinies.
Furthermore, the AI-generated content mirrors Ginzburg’s (1980) distinction between two epistemological approaches: scientific generalization and particular conjecture. While scientific generalization prioritizes abstract models, measurable data, and universal principles—stripping away individual complexities—conjectural approaches value qualitative observation, nuanced interpretation, and context-rich detail, akin to the methods of a detective or art connoisseur. Cece’s algorithmic divination embodies Ginzburg’s notion of scientific generalization, systematically reducing traditional Bazi’s relational complexity into discrete, standardized categories.
Specifically, delving into Cece app’s detailed analysis: It presents Each pillar of year, month, day, and time assigning both a heavenly (top) and earthly (bottom) component. Each combination is extracted, combined with isolated interpretation. For instance: when year+heavenly component is Friend Star(比肩星), the Cece analysis shows: the living standard in your yearly age is normal. There are not many resources and support your family can give you. You need to rely on your own efforts to achieve achievements.
The comment section beneath this reading holds more than 2,000 user responses. The top comment reads simply: “Not true. After I was born, my family’s financial situation improved.” It’s followed by over 200 replies and reposts—users affirming the same experience in different words: “Same, my birth marked the turning point in my family’s fortunes.” “My sister’s chart has this too—our parents’ business took off after she was born.” Another user states, “This is my brother’s Bazi. We moved into a bigger house the year he was born.” The commentary doesn’t stop at personal experience. Others turn to cultural references and celebrities to reinforce their dissent. “This is the same Bazi structure as Wang Sicong,” one user points out, referencing the famously wealthy son of a Chinese real estate tycoon. As the case shows, through their everyday conversations, they reclaim the particular approach—grounding abstract predictions back into personal and lived contexts. Cece’s digital forum creates a collective conversation around fate—one that is shaped by the fluid boundaries between algorithmic authority and lived reality. When an AI-generated interpretation makes a broad statement—such as suggesting that a particular Bazi characteristic correlates with financial struggle—users with the same chart element often argue back, offering counter-examples drawn from real life.
This phenomenon could be elaborated well through Badiou’ s probabilistic and epistemological model of chance versus Deleuze’s ontological dimension. (Gangle, 2010) For instance, for throwing a dice, Badiou critiques Deleuze’s notion of affirming “the whole of chance” by reducing it to a homogenized mathematical infinity, where individual events (e.g. a dice throw) dissolve into insignificance (1/∞ = 0). However, as Emanuele Severino highlights, Badiou’s critique relies on conflating epistemological probability with ontological chance. Severino contends that probabilistic reasoning (e.g. treating a die’s six faces as equally probable) masks an analytic truth—subjective ignorance of causation—as a synthetic ontological claim, thereby naturalizing probability as a tool of scientific power. In Badiou’s framework, probability models falsely equate epistemological indifference—not knowing which face will land—with ontological equivalence, erasing the material specificity of events. Deleuze, by contrast, rejects this flattening. He argues that each finite act (e.g. a dice throw or a divinatory cast) resonates with the virtual whole, generating a dynamic interplay between actual finite structures and the infinite potential of virtual multiplicities. Rather than negating individual events, Deleuze’s “whole of chance” emerges through their serial explication—each cast becomes an index of infinite possible paths. When users refute or add dimensions to algorithmic predictions with personal contexts, each comment becomes a “throw,” initiating new resonances and possibilities, emphasizing lived experience and active participation over passive acceptance of calculated outcomes. It attempts to clear the space of creation from the “givens” that return thought to cliché and impede the ramifying of events into as-yet-unknown territory. It is an index of the virtual, reopening a field of destiny to unforeseen connections.
Fragmentation and Relationality: Contesting Algorithmic Epistemology
The CeCe app’s AI-driven Bazi readings exemplify another tension: the clash between isolated listing and relational emergence in the production of divinatory knowledge. Where Badiou and Deleuze contest how chance operates, the fragmentation-relationality tension concerns how knowledge is organized. The CeCe app’s AI-generated Bazi readings present users with a series of discrete, isolated insights: a “Weak Fire Day Master” here, a “Clashing Marriage Palace” there, each diagnosis isolated into bullet-pointed predictions. This approach reflects the logic of computational parsing—breaking complex systems into digestible, algorithm-friendly units—the qualitative whole is being quantitatively separated. Yet for many users, this fragmentation clashes with the foundational premise of Bazi cosmology: that fate emerges not from isolated traits, but from the dynamic interplay of all components within a birth chart. In CeCe’s comment sections, user critiques consistently point to this epistemological dissonance. Across hundreds of threads, a recurring refrain appears:
“The AI analysis lacks relationality.”
“The heavenly (top) and earthly (bottom) components cannot be read separately.”
“You have to read the whole Bazi together.”
“You should also read it in combination with your parents’ Bazi to check for harmony.”
These objections are not only technical but ontological, challenging the app’s reduction of a holistic, correlative system into a collection of severed parts. The AI’s interpretive method reflects what Tim Ingold terms a “building block” epistemology—a worldview that constructs reality from discrete, preformed elements. Each Bazi feature (the Day Master, Hour Pillar, hidden stems) is analyzed in isolation, assigned standardized advice (“Strengthen Wood to support Fire”), and presented as a standalone data point. For instance, a user with a “Strong Earth” Day Master might receive the generic warning: “Avoid impulsive decisions; Earth favors stability.” This glosses over how Earth’s potency could be tempered or amplified by other chart elements—say, a Water-dominated Month Pillar that erodes Earth’s rigidity, or a Metal-heavy Year Branch that diverts its energy.
Users contest this fragmentation through insisting about geju (格局)—a term denoting the emergent “pattern” or “structure” formed by a birth chart’s interconnecting elements. In traditional Bazi practice, a master diviner assesses geju by tracing how elements clash, generate, or transform one another across the chart’s temporal and spatial axes. A “Weak Fire” Day Pillar might find strength through a Month Pillar rich in Wood (which feeds Fire) or be destabilized by a Water-heavy Hour Pillar (which extinguishes it). Crucially, geju is not the sum of parts but the product of their entanglement—a whole irreducible to its components. CeCe’s AI, trained on pattern recognition algorithms, fails to grasp the positionality of elements—where a feature resides in the chart’s spatiotemporal matrix (Year, Month, Day, Hour) and how it interacts with neighbors.
The disconnect between algorithmic parsing and relational logic reflects the ethics of guanxi in Han Chinese tradition and, more broadly, echoes tensions central to the ontological turn. Hall and Ames (1998) describe the focus-field model of the self as relationally embedded and situationally constituted. In this view, the self is never isolated but always entangled in a web of human relationships, where context takes precedence over individual agency. Action is shaped not by autonomous will but by one’s position within a dynamic field of roles and obligations. This contrasts with dominant Western notions of the self as an essential, bounded subject. In a similar vein, Bazi characteristics—such as the Five Elements and the Four Pillars—should not be understood as fixed properties but as relational tendencies, configurations of becoming that unfold through context, encounter, and interpretation.
Users, however, refuse this quantitative reduction. In comment threads, they piecing together relational insights from the AI’s fragmented outputs. One user, dissatisfied with her “Water Floods Fire” diagnosis, crowdsources a more nuanced reading: “My Hour Pillar has Strong Earth—can that dam the Flood?” Others respond with elemental arithmetic: “Earth controls Water, so yes! But check your shensha (神煞deities)” These exchanges reflect a pattern of “thick description”—layering context upon context to recover meaning. Here, users search for the app’s scattered parts, reassembling them into relational whole through collective labor.
Bazi divinatory systems resists mechanistic capture. Its power lies not in isolated predictions but in the relational spaces between elements—the way Wood’s growth might temper Metal’s edge, or how Water’s flow could soften Earth’s stubbornness. These interactions evoke Ingold’s metaphor of the “meshwork,” (Ingold, 2012) where life unfolds through entangled lines of growth and movement, not static nodes. The app’s AI for users, fails to trace the ever-shifting contours of the meshwork, where fate is not decoded but woven. But in that very reduction lies something unexpectedly generative. The AI’s “failure” to account for relationality is precisely what animates user responses—sparking debate, reinterpretation, and speculative layering in the comment threads. It opens a space for users to reassemble the fragments algorithmically presented, stitching them back into more holistic forms of meaning through shared cultural logic and personal narratives.
Digital Divination and Algorithmic Futures
However, the relationship between AI and divination could be more complicated, if we consider AI itself as a form of divination. AI logic and divinatory logic might not be opposite, but complementary. Davies (2024) points out that “Al achieves this divinatory task by ingesting vast quantities of data (codex) and applying Machine Learning algorithms to make connections and discover patterns in ways that exceed traditional forecasting systems (contingency).” (2024 no page number) Algorithms, like divinatory systems, draw meaning from pattern recognition and inference. But they also operate within a framework of intentional opacity. Borrowing from Latour’s (1999) concept of “black boxing,” we can understand algorithms as systems whose internal processes are either explicitly protected such as for competitive or proprietary reasons or implicitly obscured by the sheer complexity of code and neural networks. (Carlson, Wilenius, and Corliss 2023) In either case, the algorithm becomes detached, inscrutable—its predictions appear objective and impersonal, while being shielded from scrutiny, debate, and accountability (Campolo and Crawford 2019).
In China, this entanglement of prediction, governance, and opacity is made visible through the rise of algorithmic systems like the social credit initiative. Officially introduced in 2014, the Social Credit System reflects an effort to build a comprehensive framework for behavioral evaluation—linking moral conduct, legal compliance, and economic trustworthiness into a system of real-time assessment (Backer 2018). While often framed in Western media as a dystopian surveillance regime, it should be more accurately understood as a form of datafied moral governance. It aims not just to track behavior, but to shape it—to cultivate “trustworthy” subjects in service of both social order and economic efficiency (Franceschini and Loubere 2022). Across cities and corporate platforms, including Alipay’s Sesame Credit, experimental models have emerged that integrate financial activity, digital behavior, and even social relationships into predictive scoring systems that translate individual behavior into structured, feedback-based assessments of creditworthiness These are not just technical interventions, but part of a broader “social moral engineering” (Franceschini and Loubere 2022, 24) that fuses economic participation with moral cultivation.
Cece operates within this broader ecosystem. Although it positions itself as an “emotional support” app providing personalized guidance through AI-generated Bazi analysis, its internal logic is tightly aligned with a wider governance paradigm across China. Its interface design, composite scores, and tiered access to “more accurate” or “professional” fate readings mirror the light-touch behavioral modulation seen in apps like Alipay. These platforms do not mandate what users should do, but they nudge—through algorithmic scoring, upgrade options, and normative feedback loops—users to self-regulate, optimize, and align with data-driven metrics. In this sense, Cece’s divinatory outputs mirror as a governance technology, subtly directing users toward predictable and platform-aligned behaviors through the language of fate.
As Davies (2024) notes, the ambiguity of AI gains urgency in a world increasingly governed by predictive systems. Forecasting technologies now shape everything from policy decisions to financial speculation and personal opportunity. Ramey (2016), cited by Davies, describes these forecasting systems as a “disavowed form of divination”—retaining the structure of oracular prediction while masking it with the language of science and objectivity. In China, this logic is deeply embedded in the fabric of everyday life. From credit ratings and AI-driven identity verification to online finance, predictive infrastructures have become central not only to governance but to individual aspiration. As Franceschini and Loubere (2022) argue, China’s social credit initiatives are not solely instruments of state control but part of a participatory and experimental governance project, often embraced by individuals seeking inclusion and upward mobility within the formal economy. As Bear (2020) puts it, we are all now living among anticipatory tools: investment forecasts, climate models, brand futures, government plans. Within this horizon of constant calculation, individuals adapt by aligning with these systems. Can Cece users, then, fully critique the machinery they also depend on—when algorithmic fate has already become a medium of hope and self-making?
It is within this broader ecosystem that Cece becomes a microcosm of algorithmic opacity and determinism: a space where these forces are both reproduced and subverted, partially. While the platform quantifies fate through clean diagrams and AI-generated readings, users refuse to treat these outputs as conclusive. By anchoring algorithmic outputs in relational ontologies and real-life examples—manifested through concepts like geju (格局) and the sharing of personal experiences—users resist accepting the arbitrary calculation of fate produced by AI, especially when the underlying processes remain ambiguous and opaque. Instead, they assert that the power of divination lies in its capacity to foster accountability to the present—to the tangled meshwork of lives, elements, and choices that no algorithm can fully grasp. In a context where datafied, AI-driven calculation proliferates, particularly around the prediction of individual destiny, this logic is actively challenged. What emerges instead is a kaleidoscope-like reinterpretation: plural, participatory, and grounded in everyday meaning-making. However, crucially, it is this lively layer of user commentary that makes Cece both popular and socially resonant. The comment sections operate as a collective interpretive infrastructure, filling in the epistemic gaps and ambiguities left by algorithmic generation. In this way, user responses function as a form of corrective feedback— performing a kind of post-hoc calibration. Moreover, the comments could be understood as a form of user-generated data which paradoxically, feed back into the system for further optimization. In this way, user participation both challenges and sustains the algorithmic system, illustrating the deeply entangled nature of resistance and reproduction within digital divinatory infrastructures.