This article examines the fairness of recommendation systems from both consumer and provider viewpoints, discussing sources of bias, definitions of equality and equity, measurement metrics such as CGF and MMF, and proposes causal embedding models to mitigate unfairness while ensuring sustainable system performance.
The presentation explains how to detect, analyze, and leverage regional variations in overseas A/B test results to make more informed product decisions, using a systematic experimental analysis framework grounded in causal inference and online experimentation methods.
In this talk, senior researcher Dai Quanyu from Huawei Noah's Ark Lab presents his work on applying causal inference to identify and correct various biases in recommendation systems, detailing underlying theoretical frameworks, bias‑mitigation algorithms such as inverse propensity weighting and robust learning, and real‑world product deployments.
This article presents a comprehensive approach to estimating causal effects on clustered data using a double‑difference method, introduces the DiConfounder algorithm built on Rubin Causal Model extensions, details data characteristics, model assumptions, six‑step pipeline, and reports competitive results on the ACIC2022 challenge.
This article introduces the fundamentals of causal inference from observational data, explains confounding and the SUTVA assumptions, presents the do‑operator, and details four de‑confounding strategies—including RCT‑based resampling, feature‑decomposition, double machine learning, and back‑/front‑door adjustments—followed by real‑world applications in recommendation systems and resource allocation.
This article introduces the Python causalinference library, explains its core CausalModel interface and key methods for propensity‑score estimation, trimming, stratification, and various treatment‑effect estimators, and demonstrates how to interpret the resulting statistical outputs.
This article explores how causal inference techniques, including dynamic Bayesian networks and time‑series models, can be used to improve demand forecasting and replenishment strategies in inventory management, offering both theoretical concepts and practical case studies for operational decision‑making.
This article explains why causal inference is needed in data‑driven businesses, introduces its theoretical foundations from computer science, econometrics and statistics, and demonstrates how various causal modeling techniques can be used to boost user retention and content creation on the Xiaohongshu platform.
This article reviews a product‑manager’s end‑to‑end workflow for forecasting demand value and validating hypotheses, illustrating how Wallace’s scientific loop translates to business, and detailing causal‑inference techniques such as matching, DID, regression discontinuity, and instrumental variables with a real‑world case study.
This article examines the fairness of recommendation systems from both consumer and provider viewpoints, discussing sources of bias, definitions of equality and equity, measurement metrics such as CGF and MMF, causal embedding techniques, experimental results on MovieLens and Yelp, and future research directions.
This article presents a comprehensive overview of applying causal inference to Alibaba’s Business Intelligence platform, detailing fundamental concepts, the merchant intelligent diagnosis system architecture, key technologies such as hybrid causal network discovery (HCM) and deep attribution, and showcases the resulting operational impact.
This article introduces the generalized causal forest, explains its non‑parametric nonlinear construction for estimating heterogeneous dose‑response functions, compares it with existing methods, and demonstrates its experimental results and deployment in an online ride‑hailing pricing system to balance supply and demand.
Laser (Latent Surrogate Representation learning) estimates long‑term ride‑hailing market effects by inferring hidden surrogate variables from short‑term outcomes using an iVAE and inverse‑probability weighting, thereby reducing experiment cost and latency while achieving more accurate causal effect predictions than existing baselines.
Leading experts discuss how causal inference intersects with large language models, exploring opportunities, challenges, industry applications, and future research directions, while sharing personal journeys into causal reasoning and offering practical advice for practitioners.
This article presents a comprehensive overview of causal inference paradigms, the evolution of uplift and representation‑learning frameworks, and introduces the Mono‑CFR counterfactual credit‑limit model that estimates treatment effects for continuous credit limits using observational data while addressing confounding factors.
This article introduces the application of causal inference in financial user operations, outlining typical scenarios such as programmatic advertising and user outreach, discussing data and business challenges, and presenting practical implementations including propensity score matching, sample library construction, experiment design, and full‑stack uplift modeling.
This article presents a comprehensive overview of applying causal inference techniques to Tencent Weishi's cash incentive and video supply‑demand optimization, detailing business modeling, algorithmic frameworks, treatment representations, constrained multivariate causal models, experimental evaluations, and practical deployment insights.
This article presents a comprehensive overview of trustworthy intelligent decision-making, introducing a decision framework and discussing counterfactual reasoning, complex reward modeling, predictive fairness, and regulatory constraints, while highlighting practical methods and recent research advances in each sub‑area.
The paper introduces DCCL, a model‑agnostic contrastive learning framework that decouples user interest and conformity representations to address popularity bias and out‑of‑distribution challenges in recommendation systems, demonstrating significant offline and online performance gains on real‑world datasets.
This article reviews the evolution of Weibo's large‑scale recommendation technology, covering the system's business scenarios, technical roadmap, recent large model iterations, multi‑task and multi‑scenario modeling, feature engineering, consistency between recall and ranking, and emerging techniques such as causal inference and graph methods.
This article explains the fundamentals of causal inference, illustrates its distinction from correlation modeling, and demonstrates how causal techniques can be applied to limited‑resource decision problems such as knapsack optimization, ride‑hailing subsidies, and flight pricing, while also covering experimental design, popular models, evaluation metrics, and open challenges.
This article presents a data‑science‑focused guide on using causal inference and uplift models to drive overseas ad targeting and user‑operation decisions in games, covering audience selection, privacy‑aware exposure correction, bid optimization, experiment design pitfalls, network effects, and practical recommendations.
This article reviews recent developments in causal representation learning, explaining why causal reasoning is essential, describing methods for i.i.d. data, time‑series, and multi‑distribution scenarios, and illustrating applications such as domain adaptation, video analysis, and financial data with numerous examples and visualizations.
The article explains JD’s Explore & Exploit (EE) module, its bias‑related challenges, the iterative optimization loop, model debiasing techniques for position and popularity bias, personalized bias modeling, causal inference methods, online AB results, and offline evaluation metrics, highlighting significant improvements in search diversity and efficiency.
This article explains the concept of Granger causality, outlines the four possible test outcomes, and provides a step‑by‑step Python implementation using statsmodels to determine whether one time‑series statistically causes another.
This article provides a comprehensive overview of causal inference, explaining its definition, the distinction between correlation and causation, classic pitfalls such as Simpson's paradox, key metrics like ATE and ATT, experimental designs, bias mitigation techniques, and practical case studies from content platforms and the Titanic dataset.
After attending the DataFun causal inference summit, this article outlines why causal analysis matters, its typical use cases, practical challenges, its relationship with A/B testing, and how it integrates with machine learning and deep learning to improve decision‑making and model robustness.
This article presents the JD Retail search data‑science team's practical approaches to e‑commerce search, covering the scene’s unique data characteristics, order attribution methods, AB experiment design, causal‑inference frameworks, variance‑reduction techniques, quasi‑experimental evaluations, and metric design for traffic distribution, all illustrated with real‑world examples and visualizations.
The paper tackles low-response A/B tests by applying instrumental-variable techniques and optimized propensity-score matching, showing that IV methods recover treatment effects for compliant users and that a refined PSM pipeline dramatically boosts lift detection, turning previously non-significant results into statistically significant business insights.
AB TestingCausal InferencePropensity Score Matching
This presentation introduces two recent works from Ant Group that improve causal inference by explicitly using historical control data to reduce selection bias and by fusing heterogeneous multi‑source data, describing the GBCT and WMDL methods, their theoretical foundations, experimental results, and practical applications in finance.
This article reviews the two main streams of causal inference—potential‑outcome (Rubin) models and structural causal (Pearl) diagrams—covers classic techniques such as A/B testing, instrumental variables, matching, difference‑in‑differences, synthetic controls, matrix completion, heterogeneous treatment effect estimation, and discusses modern machine‑learning‑based approaches and causal discovery algorithms.
This article introduces the fundamental ideas of causal inference, explains how it differs from correlation‑based machine learning, discusses the role of confounders, and showcases practical implementations in financial services such as offer optimization, uplift modeling, and decision‑making pipelines.
This article surveys the development of artificial intelligence from symbolic and connectionist perspectives, covering deductive and inductive reasoning, multimodal and cross‑modal inference, knowledge‑graph reasoning, text and visual understanding, and their applications in causal inference, dialogue consistency, and security vulnerability analysis.
This article summarizes the motivations behind causal inference, its typical business applications such as intelligent decision‑making and prediction, the practical challenges of validation and data, and its relationship with A/B testing, machine learning, and deep learning, providing a concise overview for newcomers.
The article describes Tubi’s internal experimentation platform, explaining how traditional user‑group A/B tests can suffer from network interference and how Switchback experiments—time‑window based designs—address these issues, detailing their implementation, statistical methods, and the practical challenges overcome.
This article introduces causal inference methodologies tailored for game scenarios, discusses the challenges of offline inference on massive data, and presents three distributed solutions—low‑complexity propensity‑score matching, robust double‑robust estimation, and panel difference‑in‑differences—along with their implementation details and performance insights.
Understanding why causal relationships matter reveals the limits of predictive machine learning, introduces counterfactual reasoning, explains potential outcomes, treatment effects, bias, and how to distinguish correlation from causation using simple examples like tablet distribution in schools.
The article explains how causal inference, particularly Propensity Score Matching, can be used to control confounding factors and accurately estimate the incremental value of a marketing campaign when randomized experiments are infeasible, illustrating the method with a real Ctrip project case study.
This article explains how adding covariates that are not confounders can reduce outcome variance and improve causal inference, while controlling for variables that only predict treatment may introduce selection bias and inflate estimation error.
Causal Inferencecontrol variablesobservational study
This article explains how linear regression can be used to estimate average causal effects, handle bias, and draw valid conclusions from both randomized experiments and observational data, while illustrating the theory with concrete examples and visualizations.
This article introduces causal graphical models as a language for reasoning about cause‑and‑effect, explains key concepts such as conditional independence, colliders, back‑door paths, confounding and selection bias, and shows how to identify and adjust for bias using simple visual examples.
This article explains the fundamentals of graph networks, their use in Kuaishou’s short‑video ecosystem for community detection, trend‑driven content creation, and causal inference, and describes how key network nodes and frequency‑illusion effects are leveraged to boost user engagement and content virality.
This article explains why randomised experiments are the gold standard for turning correlations into causal claims, illustrates their use in evaluating online versus face‑to‑face learning, and discusses ideal experimental design, assignment mechanisms, and key take‑aways for causal inference.
This article explains why causal inference matters beyond prediction, introduces potential outcomes notation, demonstrates how bias separates correlation from causation, and outlines the conditions under which observed differences can be interpreted as true causal effects.
This article presents a data analyst’s walkthrough of player churn analysis for the battle‑royale game Naraka: Bladepoint, illustrating how survival analysis, epidemiological experiment designs, and econometric causal inference methods can uncover systemic and event‑driven attrition and guide more effective game‑operation strategies.
This article explains AB testing—from its biomedical origins and online adoption to its types, three essential components, core values of causal validation and quantitative growth, and key characteristics of pre‑evaluation and parallelism—providing a comprehensive guide for data‑driven product optimization.
CausalMTA introduces a causal‑inference‑based multi‑touch attribution framework that eliminates static user‑attribute bias through variational recurrent auto‑encoder re‑weighting and dynamic preference bias via a gradient‑reversal‑layer‑augmented RNN, producing an unbiased conversion predictor whose Shapley‑value credits outperform existing methods on synthetic, Criteo, and Alibaba datasets.
This article explains the fundamentals of A/B testing, how Netflix designs experiments with random sampling and assignment, evaluates metrics to infer causal impact, and uses these insights to guide product decisions and improve member engagement.
This article introduces Feizhu's advertising diagnosis platform and explains how recent causal inference techniques, especially the NO TEARS algorithm and Bayesian‑network‑based attribution, are applied to identify the root causes of performance fluctuations across the ad delivery funnel, improve diagnostic accuracy, and guide optimization decisions.
Ad AttributionAdvertising DiagnosisBayesian network
The article explains Propensity Score Matching as a causal inference method, detailing treatment effect concepts, required assumptions, score estimation, various matching algorithms, SQL implementation, quality assessment metrics, and how to estimate ATT using Difference-in-Differences, while outlining workflow steps, trade-offs, and alternatives.
This article explains observational causal inference fundamentals, compares methods like propensity score matching, Bayesian causal graphs, and difference‑in‑differences, and demonstrates their practical application in evaluating the business impact of Taobao's 3D sample rooms.
This article explains how Hello Mobility uses tree‑based causal inference and uplift modeling to improve smart subsidy efficiency in hotel marketing, covering background, uplift methods, custom split criteria, offline AUUC evaluation, online deployment, and future research directions.
This article explores how combining causal inference with machine learning can uncover subtle correlations in large datasets, detailing user growth metrics, propensity‑score matching, causal recommendation models, heterogeneous treatment effect analysis, and practical strategies for improving retention and activity in recommendation systems.
This talk presents the evolution of QQ Music's ranking system, detailing background challenges, user‑perception modeling, multi‑objective and causal learning to mitigate the Matthew effect, long‑tail content support, cross‑domain recommendation, and module personalization for diversified traffic, concluding with future research directions.
The article reviews the limitations of association‑based AI, explains the two main causal inference paradigms, introduces differentiable causal discovery, and shows how these ideas address out‑of‑distribution challenges and stable learning in recommendation systems, citing recent research.
The article introduces causal inference for observational business data, contrasts methods that require observed confounders with instrumental-variable techniques that can address unobserved confounding, explains the three core IV assumptions plus homogeneity or monotonicity, illustrates the Wald estimator, warns about weak instruments, and urges careful application.
This talk explores how causal inference techniques are used to identify and mitigate various biases in Kuaishou's recommendation pipeline, covering background theory, recent research advances, practical implementations for popularity and video completion debiasing, and reflections on challenges and future directions.
The paper applies causal inference and tree‑based uplift modeling to identify coupon‑responsive riders, using T‑, S‑, and X‑Learners as well as a proprietary Treelift model that directly optimizes per‑user utility, achieving a 4.7% lift over manual rules and 2.3% over prior response models.
This article presents a comprehensive overview of how causal inference techniques are applied to identify and correct various biases in Kuaishou's recommendation pipeline, covering background theory, recent research, practical implementations such as popularity debias, causal embedding decoupling, and video completion‑rate debias, along with experimental results and future challenges.
This article explains why causal inference is essential for decision‑making, contrasts it with pure prediction, outlines the four DoWhy steps (modeling, identification, estimation, refutation), and demonstrates a case study on road freight accidents using DoWhy and EconML with code examples and results.
This article details Tencent's application of uplift modeling to optimize QQ Browser push notification volume, covering push system characteristics, causal analysis challenges, a refined S‑learner with metric learning, and resulting DAU improvements, while also addressing practical Q&A on uplift techniques.
This article introduces the fundamentals of causal inference with observational data, explains confounding and collider structures, compares observational and experimental approaches, discusses challenges such as Simpson’s paradox, and presents Tencent’s quasi‑experimental applications including DID, regression discontinuity, and uplift modeling.
This article provides a comprehensive overview of causal inference on observational data, explaining confounding and collider structures, experimental solutions, the differences between observational and experimental data, challenges such as Simpson's paradox, and detailed Tencent case studies using DID, regression discontinuity, and uplift modeling to guide practical analysis.
The article examines how low‑level recommendation‑algorithm improvements in information‑flow feeds can boost user retention, LTV and overall growth by addressing cold‑start challenges, survivor bias, and causal inference through personalized ranking, ecosystem construction, and multi‑task learning.
The presentation offers a comprehensive overview of online experiment fundamentals, design variations, and real-world case studies from Tencent Lookpoint, emphasizing hypothesis validation, causal analysis, best practices, and actionable recommendations for improving product growth and decision‑making.
This article presents a comprehensive overview of online experiment design and analysis, covering basic definitions, AB testing principles, complex experiment types, real-world case studies from Tencent's information flow platform, and practical guidelines for reliable experiment evaluation and product decision‑making.
Using a road freight accident prediction example, the article warns that interpreting predictive model explanations as causal effects can be misleading, explains when such models may answer causal questions, demonstrates SHAP analysis on an XGBoost model, and recommends causal inference tools like ecoml for reliable effect estimation.
This article explains how Kuaishou applies causal inference frameworks, such as Rubin's potential outcomes and Pearl's causal graphs, together with machine‑learning techniques like double‑machine learning, causal forests, and meta‑learners to evaluate product features, recommendation strategies, and user behavior under complex network effects in live streaming.
This article analyzes Kuaishou's live‑streaming ecosystem, detailing causal‑inference frameworks, observational and experimental techniques such as DID, double machine learning, causal forests, uplift meta‑learners, and complex experiment designs like dual‑sided and time‑slice rotation to evaluate product and recommendation strategies.
This article presents Dr. Jin Yaran’s comprehensive overview of causal inference challenges, frameworks, and practical case studies—including DID, double machine learning, causal forests, and meta‑learners—applied to Kuaishou’s live‑streaming product, and discusses complex experimental designs such as bilateral and time‑slice experiments.
This article explains how Hulu leverages machine learning and AI techniques such as ad targeting, inventory prediction, flow matching, conversion rate optimization, causal inference, and shared‑account detection to improve the efficiency, effectiveness, and revenue of its video advertising ecosystem.
The article explains offline sampling for AB testing, detailing why it is needed, the main challenges of limited sample size, population heterogeneity, and non‑random interventions, and presents variance‑reduction, stratified sampling, IPW, and matching methods to address these issues.
Offline sampling in A/B testing assigns experimental units such as users or tags before a trial begins, but suffers from limited sample size, high heterogeneity, and non‑random allocation, which can be mitigated by variance‑reduction methods like CUPED, stratified sampling with inverse‑probability weighting, and matching approaches including propensity‑score matching.
This article explains the concept, history, and various forms of online experiments such as AB testing, ABn, AA, and multivariate tests, highlighting their role in causal inference, value evaluation, risk control, and product optimization within modern internet businesses.
Online experiments, rooted in biomedical randomized controlled trials, have become essential for internet businesses to achieve data‑driven growth by providing causal inference, quantifying value, and managing risk through various designs such as AB, ABn, AA, multivariate and quasi‑experimental tests.
The article introduces causal inference for ride‑hailing businesses, explaining the difference between causality and correlation, common misconceptions, and how randomized experiments and observational techniques like propensity‑score matching can quantify effects of actions such as coupons, driver assignments, and platform growth decisions.
Causal Inferencebusiness decisioncorrelation vs causation
This article explains how causal inference techniques are applied to model the uplift effect of ad placement in Alibaba's Fliggy search CPC advertising, transforming budget allocation into a multi‑objective optimization problem and describing practical control methods, feature engineering, sample re‑sampling, model designs, uplift evaluation, and future research directions.
This article introduces the Continuous Causal Forest, a novel uplift modeling approach that expands binary treatment effect estimation to handle multivariate and continuous treatment variables, demonstrates its construction, evaluates its performance on ride‑hailing pricing strategies, and discusses its advantages, limitations, and future research directions.
This article examines how low‑level recommendation algorithm techniques can upgrade content‑feed systems to boost user growth, covering problem analysis, growth factors, personalization upgrades, cold‑start mechanisms, bias mitigation via causal inference, and utility‑driven user profiling.
This article explains how data analysts can use experimental and observational research methods, including randomized controlled trials, quasi‑experiments, difference‑in‑differences, regression discontinuity, synthetic control, and Bayesian structural time‑series, to assess the causal impact of product and marketing changes on business metrics.
This article introduces uplift (incremental) modeling as a causal inference technique for intelligent marketing, explains its mathematical formulation, compares response and uplift models, describes various modeling approaches such as two‑model, one‑model, and label‑transformation methods, outlines evaluation metrics like Qini and AUUC, and demonstrates practical deployment in a real‑world real‑estate platform.
The article explains how Alibaba Entertainment leverages causal inference and uplift modeling to build unbiased user‑cf recommendation algorithms that model user states and upgrade personalized distribution, achieving significant click‑through and re‑activation gains for long‑video services like Youku.
The talk explains uplift modeling for intelligent marketing, showing how causal lift predictions—derived from randomized experiments using two‑model, one‑model, or tree‑based methods—identify truly responsive users, evaluate performance with AUUC/Qini, and were applied to Taopiaopiao’s coupon allocation via knapsack optimization, highlighting challenges and future directions.
The article explains how data products create business value through application, data, and algorithm layers, using examples like 5G infrared temperature screening and ImageNet, and discusses the roles of experimental design, causal inference, and reinforcement learning in building effective AI‑driven strategies.
Machine learning models often suffer from poor explainability and unstable predictions due to reliance on spurious correlations, but by applying causal inference to separate true causal relationships from confounding and selection bias, a causal‑constrained stable learning framework can achieve more interpretable and robust predictions across varying data distributions.
This article details how Hulu leverages machine learning and AI techniques—including ad targeting, inventory prediction, conversion rate optimization, causal inference, and real‑time bidding—to improve ad efficiency, user experience, and revenue across its video streaming platform.