The review was deliberately broad in scope. Rather than a narrow systematic review focused on a single population or intervention, the authors employed PRISMA-ScR guidelines to capture the breadth and diversity of evidence across three continents. Starting from 1,387 records identified in Scopus, Web of Science, and Google Scholar, the team screened out 81 non-English documents, 775 conference papers, books, theses, and other ineligible formats, and 439 irrelevant or inaccessible studies, ultimately retaining 92 sources: 77 original journal articles, 7 review articles, and 8 official reports. Fifty-eight studies were quantitative, 22 qualitative, 4 mixed-methods, and 8 were official reports. Geographically, Asia contributed 30 studies, Africa 27, South America 9, with the remainder drawn from Europe, North America, Australia, and global datasets. The authors note that roughly 83% of the included literature appeared between 2020 and 2024, signaling rapidly accelerating scholarly interest in digital adoption as a development lever aligned with the 2030 Sustainable Development Goals.

To organize this heterogeneous evidence, the researchers combined two established theoretical engines. The Unified Theory of Acceptance and Use of Technology, in its consumer-focused UTAUT2 extension, explains adoption through performance expectancy, effort expectancy, social influence, and facilitating conditions, while the Sustainable Livelihood Framework describes how households mobilize physical, social, financial, human, and natural capital to pursue livelihood strategies. Gender was modeled as a mediating factor, drawing on Naila Kabeer’s empowerment framework—resources, agency, and achievements—and on Kimberlé Crenshaw’s theory of intersectionality, which highlights that benefits and burdens from technology are not distributed uniformly across populations. Together, these lenses allowed the review to connect the micro-level decision to buy a phone with macro-level outcomes such as income diversification and multidimensional poverty.

The headline finding is a starkly divided digital geography. As of 2023, only Africa and the Arab states still relied predominantly on 3G networks while other regions had moved to 4G and 5G. Africa has the lowest mobile phone adoption rate at 63%, meaning 37% of its population remains without a handset, followed by South America at 72%. Mobile internet penetration shows the same pattern: only about 37% of Africans use mobile internet, compared with 65% in South America, while fewer than 10% of Europeans lack access to either. Perhaps most telling is the gap between phone ownership and actual internet use—26 percentage points in Africa, versus 7 in South America and 9 in Asia—a divide the authors attribute to weak 4G and 5G infrastructure that keeps users tethered to basic phones and constitutes a second-level digital divide. Rural-urban disparities compound the problem: the ratio of urban to rural mobile internet use stands at 2.5 in Africa, 1.5 in the Asia-Pacific region, 1.22 in the Americas, and just 1.04 in Europe, meaning African rural users face the widest connectivity chasm in the Global South.

Gender emerges as one of the review’s most consequential threads. The gender gap in mobile ownership is 15% in South Asia, 13% in Africa excluding North Africa, and a strikingly low 1% in South America. In mobile internet use, South Asia narrowed its gender gap from 41% in 2017 to 31% in 2023, largely credited to initiatives in India improving women’s online access, while Sub-Saharan Africa’s gap barely budged, sliding only from 34% to 32% over the same period. The evidence goes deeper than statistics: married women show significantly lower adoption and use of mobile devices than married men, and women are more likely than men to depend on borrowing someone else’s phone to access information—a pattern the authors link to patriarchal control over resources and the confinement of women to unpaid household labor. Expanding women’s access, they argue, would rebalance household bargaining power, expand skills and social networks, and open employment pathways, making gender-sensitive digital inclusion a central poverty intervention rather than a peripheral concern.

Along agricultural value chains, the review documents how phones function as production and marketing infrastructure. In Tanzania’s pastoral communities, mobile phones facilitated communication vital for cattle farming; in South Africa, they gave smallholders access to extension advice, money transfers, pasture management tips, weather updates, and input application guidance; in China, farmers used mobile apps for farm management, market information, and extension training that boosted production and income. Yet intensity of use varies enormously between and within countries. Farmers in China, India, Kenya, and South Africa deploy advanced applications for production and marketing, while counterparts in Ethiopia and Peru primarily use phones for calls and messaging. Large-scale farmers consistently out-adopt smallholders in advanced applications, a gap the review attributes to digital skill deficits, affordability constraints, language barriers, and patchy network coverage. Intermediaries remain another bottleneck: rural African phone owners still depend heavily on middlemen who compress profit margins, whereas Asian e-commerce systems increasingly connect producers directly to consumers, and Latin American farmers have used phones to raise incomes by trimming transaction chains.

Financial inclusion is where Africa leads the world. Sub-Saharan Africa registers the highest mobile money adoption of any region, and the evidence reviewed shows measurable welfare consequences. In Kenya, digital financial services have improved rural poor livelihoods, raised per capita incomes, and lifted households out of poverty; classic research on Kenya’s mobile money revolution found it significantly enhanced households’ ability to share risk by slashing transaction costs, while other studies document its role in remittance flows across domestic and international borders. Comparative assessments rank mobile money as easier, safer, more trustworthy, more convenient, faster, and cheaper than traditional banking, and its reach into unbanked rural communities is credited with stimulating economic participation. By contrast, mobile money remains underutilized in Asia and especially South America, depriving marginal populations in infrastructure-poor areas of a proven tool—an asymmetry the review urges policymakers to correct through targeted promotion.

The poverty picture, however, is not uniformly rosy, and the review is candid about the reversals. Studies in selected Sub-Saharan African countries found that internet-enabled mobile phones were associated with increased poverty, and some economists argue that ICT-driven sectors reward skilled workers while trapping less-skilled laborers in low-paid jobs, deepening income inequality. Household budgets can suffer too, as communication expenditures divert money from food and necessities, and some research even links expanding phone coverage to increased violent conflict. On the livelihood side, phones open new economic opportunities, reduce risk exposure, lower transaction costs, and connect farmers to end users in India, South America, and beyond, but poor network coverage, unreliable electricity, low digital literacy, and constrained incomes blunt these gains, particularly in rural areas where over half of Africa’s population lives.

The authors’ conclusions amount to a policy agenda calibrated by region. For Sub-Saharan Africa, they prioritize expanding digital infrastructure, deepening mobile-based agricultural extension, building digital skills, and dismantling the affordability and gender barriers that keep women and smallholders offline. In South America, where empirical evidence remains thin and largely limited to English-language literature, they call for research and funding on mobile money and e-commerce adoption among women, low-income, and rural households. In Asia, attention should turn to gender-based and socioeconomic inequalities in access and to rural entrepreneurship. Across all three regions, they recommend regulatory support, targeted subsidies for underserved groups, digital platforms matched to local skill profiles, awareness campaigns, and electricity expansion. Their closing warning is stark: unless developing countries act decisively to close the mobile divide now, each new wave of technology will only widen the gap between individuals, businesses, countries, and continents.

Subject of Research: Mobile-based digital technology adoption, gender, livelihood diversification, and poverty reduction in the Global South

Article Title: Mobile-based digital technology adoption, gender, livelihood diversification, and poverty reduction in the Global South: a scoping review

Article References: Bule, D. L., Ntuli, H., Wale, E., & Nketiah, P. (2026). Mobile-based digital technology adoption, gender, livelihood diversification, and poverty reduction in the Global South: a scoping review. SN Social Sciences, 6(10), Article 459. https://doi.org/10.1007/s43545-026-01749-2

Image Credits: AI Generated

DOI: 10.1007/s43545-026-01749-2

Keywords: mobile technology adoption, digital divide, Global South, poverty reduction, livelihood diversification, gender gap, mobile money, e-commerce, smallholder farmers, digital infrastructure, financial inclusion, scoping review

The research, conducted by Vijay H. Kalmani, Amol C. Adamuthe, and Pooja Bagane at institutions in Maharashtra, India, introduces a system called Sentiment-Aware Redundancy-Minimized Maximum Marginal Relevance, or SA-RMMR. Unlike the large generative language models that have dominated recent headlines, SA-RMMR is an extractive summarizer: it selects actual sentences written by real customers rather than generating new text. That design choice is deliberate. Because the summary sentences come directly from the review corpus, the framework cannot hallucinate claims that no reviewer ever made, a persistent weakness of abstractive models such as BART, PEGASUS, and FLAN-T5.

Technically, the pipeline begins by grouping all reviews belonging to a single product, since the goal is a product-level digest rather than a summary of one individual review. The raw text is cleaned, tokenized, lemmatized, and split into candidate sentences. Each sentence is then converted into a dense semantic vector using Sentence-BERT, specifically the all-MiniLM-L6-v2 model, which allows the system to measure how semantically close a sentence is to the overall meaning of the entire review set. The framework computes a centroid vector, essentially the average of all sentence embeddings, and scores each candidate by its cosine similarity to that centroid. Sentences near the centroid capture the collective voice of the reviewers.

Where SA-RMMR departs from conventional extractive methods is in the additional signals it folds into the ranking. Product aspects are extracted automatically using part-of-speech tagging with the spaCy language model, retaining nouns and proper nouns as candidate aspect terms, which are then weighted with TF-IDF restricted to that aspect vocabulary. A sentence that mentions frequently discussed and statistically important aspects, such as battery life, shipping, or build quality, earns a higher aspect coverage score. Sentiment is handled by the VADER analyzer, which assigns each review a polarity label of positive, negative, or neutral. The framework then computes how closely the sentiment distribution of a candidate summary matches the distribution of the original reviews, so that a product loved by seventy percent of buyers but panned by thirty percent does not get summarized as universally adored.

The final ingredient is redundancy control, adapted from the classical Maximum Marginal Relevance algorithm used in information retrieval. At each selection step, the framework scores every remaining candidate sentence as a weighted combination of semantic relevance, aspect coverage, and sentiment alignment, minus a penalty proportional to the maximum cosine similarity between that sentence and anything already chosen. Reviewers tend to phrase identical complaints in endlessly varied ways, and a naive relevance-based selector will happily include five near-duplicate sentences about slow delivery. The redundancy penalty suppresses that behavior, forcing each additional sentence to contribute genuinely new information. The authors also add implementation refinements, including semantic centrality estimation, adaptive retrieval confidence, reward mixing, and a balance bonus, which they tuned through a two-stage validation procedure and an Optuna-based hyperparameter search using a Tree-structured Parzen Estimator over 150 trials.

Equally important is how the summaries are judged. The team argues that ROUGE-style lexical overlap metrics, the industry standard for summarization evaluation, are poorly suited to opinion mining, because two summaries can overlap heavily in wording while differing dramatically in which product features they cover or which sentiments they preserve. To address this, the researchers propose a Composite Quality Index, or CQI, that jointly measures semantic relevance, aspect coverage, sentiment alignment, diversity, and redundancy, with equal weights assigned to each dimension in the reported experiments. A high CQI score means a summary is representative, feature-complete, emotionally faithful, varied, and non-repetitive all at once, a far stricter standard than word overlap alone.

Experiments on Amazon product review datasets, drawn from a publicly available processed repository, compared SA-RMMR against a broad set of baselines: the simple Lead-k heuristic, the graph-based TextRank and LexRank algorithms, a centroid-based SBERT extractor, and three zero-shot transformer models, BART, PEGASUS, and FLAN-T5. All methods received the same pre-processed inputs and the same summary length constraint of four sentences, and results were averaged over five independent runs with different random seeds. The proposed framework achieved the highest aspect coverage, 0.835 in the abstracted benchmark configuration, and the highest sentiment alignment, 0.719, while maintaining a competitive BERTScore-F1 of 0.825. Although a centroid-based SBERT baseline edged out SA-RMMR on raw ROUGE-L, its composite quality was lower because it covered fewer aspects and repeated itself more.

The transformer baselines fared considerably worse on the task-specific criteria. BART reached a CQI of 0.444, PEGASUS 0.432, and FLAN-T5 0.411, with the authors attributing the gap to weaker preservation of domain-specific aspects and less accurate sentiment retention despite the models’ fluent generation. Statistical significance testing using paired t-tests or Wilcoxon signed-rank tests confirmed that SA-RMMR’s advantages over the transformer models were significant at the p < 0.05 level, and Cohen’s d effect sizes indicated large practical gains in semantic quality relative to BART and FLAN-T5. Ablation studies reinforced the architecture’s logic: removing SBERT-based semantic representation caused the largest drop in composite quality, followed by removing aspect awareness and adaptive retrieval, while sentiment alignment, redundancy control, and centrality each contributed smaller but consistently positive effects.

A qualitative case study on a representative product illustrated the trade-offs vividly. TextRank achieved broad aspect coverage but at the cost of the highest redundancy, effectively telling the reader the same thing several times. The abstractive transformers produced short, readable summaries that skipped many product characteristics entirely. SA-RMMR landed between the extremes, covering hardware, usability, reception, camera, and sound quality while driving its redundancy score to zero and retaining both positive and negative opinions. The authors are candid about limitations: the framework depends on the quality of VADER sentiment labels and may stumble on sarcasm, irony, or figurative language; automatically built aspect vocabularies can miss rare or implicit features; the weighting coefficients were set empirically; and no human evaluation was included. Future work, they suggest, could add sarcasm-aware sentiment modeling, LLM-based aspect extraction, multilingual support, credibility estimation, and human-centered evaluation protocols.

The broader significance lies in what the study says about how opinion summarization should be built and measured. As e-commerce platforms, recommendation engines, and even market researchers increasingly rely on automated digests of user-generated content, summaries that silently amplify majority sentiment or erase minority concerns can distort purchasing decisions at scale. By demonstrating that an interpretable, extractive pipeline, combining sentence embeddings, linguistic aspect extraction, sentiment distribution matching, and redundancy-aware ranking, can outperform much larger generative models on the criteria that actually matter to readers, the SA-RMMR framework makes a case for efficiency and transparency over raw model size. Its Composite Quality Index, meanwhile, offers the field a template for evaluating summaries the way consumers actually use them: as faithful, balanced, and complete portraits of collective opinion rather than as lexical puzzles solved for a benchmark.

Subject of Research: A sentiment-aware, redundancy-minimized extractive summarization framework for aspect-aware product review summarization in e-commerce.

Article Title: Sentiment-guided semantic ranking framework for aspect-aware product review summarization

Article References: Kalmani, V. H., Adamuthe, A. C., & Bagane, P. (2026). Sentiment-guided semantic ranking framework for aspect-aware product review summarization. Discover Artificial Intelligence, 6(1), Article 1125. https://doi.org/10.1007/s44163-026-02176-1

Image Credits: AI Generated

DOI: 10.1007/s44163-026-02176-1

Keywords: product review summarization, sentiment analysis, natural language processing, Sentence-BERT, maximum marginal relevance, aspect extraction, VADER, extractive summarization, e-commerce, Composite Quality Index, text mining, machine learning

The study, conducted by Xiaochen Sun of the School of Product Design at Shanghai Art and Design Academy, examined seven factors that previous literature has linked to how customers perceive an online store: communication quality, usability, information quality, customer service, website design, security, and privacy. While each of these dimensions has been studied individually, the relative weight they carry when measured together against customer satisfaction has remained poorly understood. Disentangling those relationships matters because platforms invest heavily across all seven fronts, often without knowing where each additional unit of design effort yields the greatest return in customer goodwill.

To build the dataset, the researcher carried out a quantitative survey of 400 people, each of whom had made at least one online purchase within the previous four months. Respondents answered a structured questionnaire in which every item was rated on a seven-point Likert scale, a standard psychometric instrument that asks participants to express degrees of agreement, from strong disagreement to strong agreement. Restricting the sample to recent purchasers was a deliberate methodological choice: it ensured that judgments about usability, communication, and design were grounded in fresh, first-hand experience rather than hazy recollections of past shopping episodes.

The analysis pipeline followed the conventions of survey-based behavioral research. All computations were performed in SPSS, a widely used statistical software package. The researcher first tested reliability, confirming that the questionnaire items measuring each construct produced internally consistent responses. The result was striking: Cronbach’s alpha, the most common reliability coefficient, reached 0.959, a value far above the conventional 0.70 threshold and indicative of exceptional consistency across the measurement items. Validity checks followed, establishing that the questionnaire genuinely captured the constructs it was designed to measure.

With the measurement instrument validated, the study turned to inferential statistics. Correlation analysis was used to explore the raw associations between each UX factor and customer satisfaction. An analysis of variance, or ANOVA, was then applied to test whether the observed differences across groups were statistically meaningful. Finally, the centerpiece of the analysis was a multiple linear regression model, a technique that estimates how much each predictor variable contributes to an outcome while holding the other predictors constant. This is the crucial step that separates genuine drivers of satisfaction from variables that merely correlate with it through their relationship with other factors.

The regression model proved statistically significant, with an F statistic of 4.548 and a p value below 0.001, meaning the probability that the observed pattern arose by chance alone is vanishingly small. Within that model, four factors emerged as significant predictors of customer satisfaction: communication, usability, website design, and privacy. In practical terms, shoppers were most satisfied when platforms communicated clearly and responsively, when interfaces were easy to navigate, when visual and structural design felt polished, and when the handling of personal data inspired confidence rather than suspicion.

Just as revealing were the factors that failed to reach significance. Information quality, customer service, and security did not turn out to be significant predictors of satisfaction in this dataset. The finding does not mean these dimensions are irrelevant to e-commerce success; rather, it suggests that they may function as baseline expectations. Modern shoppers may simply assume that product information will be accurate, that help will be available when needed, and that transactions will be protected, so variation in these areas contributes little additional satisfaction once that baseline is met. Satisfaction, in this reading, is won or lost on the experiential frontiers of communication, usability, design, and privacy rather than on the hygiene factors shoppers take for granted.

The privacy result deserves particular attention in an era of intensifying scrutiny of how digital businesses handle personal information. That privacy registered as a significant driver of satisfaction, alongside more classically experiential factors, indicates that data stewardship has crossed over from a compliance concern into the core of the customer experience. For platform operators, the implication is that transparent data practices and respectful handling of personal information are not merely defensive measures against regulators but active contributors to how satisfied customers feel about their shopping journey.

For the e-commerce industry, the study offers a prioritized roadmap. Platforms operating under finite design and engineering budgets can direct resources toward the four significant factors, refining how they communicate with shoppers, streamlining navigation and interaction flows, investing in the aesthetic and structural quality of their sites, and making privacy protections visible and trustworthy. The author concludes that these UX aspects are crucial for e-commerce providers to manage if they want to keep customers satisfied and their platforms effective. As competition in online retail intensifies and switching costs for consumers approach zero, the difference between a thriving marketplace and an abandoned shopping cart may come down to precisely these four dimensions of the user experience.

Subject of Research: The impact of user experience design factors on customer satisfaction in e-commerce platforms

Article Title: The impact of user experience design on customer satisfaction in E-commerce platforms

Article References: Sun, X. (2026). The impact of user experience design on customer satisfaction in E-commerce platforms. Journal of Ambient Intelligence and Humanized Computing. https://doi.org/10.1007/s12652-026-05121-3

Image Credits: AI Generated

DOI: 10.1007/s12652-026-05121-3

Keywords: user experience, e-commerce, customer satisfaction, usability, website design, privacy, communication, online shopping, regression analysis, survey research, SPSS, consumer behavior

The core problem the research addresses is well known among practitioners. Browsing a product, clicking on it, adding it to a cart and finally purchasing it are actions of very different kinds, each carrying its own weight and its own temporal rhythm. When recommendation systems treat these signals in isolation, or simply pool them without accounting for their structural differences, the resulting picture of user interest becomes both incomplete and inconsistent. A user who browses dozens of laptops but purchases none is signaling something quite different from a user who browses two and buys one, and a system that cannot distinguish between these patterns will make unstable, inaccurate suggestions. Zhu’s framework responds by performing unified and standardized modeling of these diverse behavioral signals, so that each action type is represented in a form that can be compared and combined with the others.

Central to the method is the idea that preferences are not static. The framework constructs differentiated feature representations under temporal semantic constraints, meaning that the timing and sequence of behaviors shape how those behaviors are encoded. Preference intensity, how strongly a user leans toward a product category, and temporal dynamics, how that leaning shifts over time, are captured as distinct but related properties of the representation. This dual emphasis allows the model to reflect the reality that a user’s interest in, say, running shoes in January may fade by March, while their interest in a different category may surge in the interim. By encoding these dynamics explicitly rather than relying on aggregate counts, the method seeks to preserve the freshness and decay of interests that traditional collaborative filtering approaches often flatten away.

The fusion stage of the framework is where the technical architecture becomes most distinctive. Rather than concatenating features from different behavior sources or averaging their predictions, the method introduces a collaborative fusion mechanism operating within a shared representation space. In this space, multi-source behavioral information is jointly modeled so that the resulting user interest embeddings are structurally consistent, meaning they share a common geometry across behavior types, and informationally complementary, meaning each source contributes what the others lack. A purchase history, for example, is sparse but highly reliable, while browsing data is abundant but noisy; the fusion mechanism is designed to let the reliability of one signal compensate for the noise of another without allowing the noisy signal to overwhelm the trustworthy one.

Once these fused interest embeddings are generated, the framework performs user-item matching directly in the learned representation space. Items are embedded alongside users, and recommendations are produced by measuring the proximity between a user’s fused interest vector and candidate item vectors. Because the interest representation already accounts for multiple behavior types and their temporal structure, the matching step inherits that richness, and the authors argue this is what enables the improved accuracy and stability observed in their experiments.

The evaluation was conducted on the publicly available Amazon multi-behavior dataset, a widely used benchmark that records browsing, adding to cart and purchasing actions alongside clicks. Under the adopted evaluation setting, the proposed approach achieved a Precision@10 of 0.412, a Recall@10 of 0.356 and an NDCG@10 of 0.437. Precision@10 measures the fraction of the top ten recommended items that were actually relevant, Recall@10 captures how many of the user’s relevant items appeared in the top ten, and NDCG@10 rewards systems that place the most relevant items near the top of the ranked list. Together, these metrics indicate that the fused representations produce rankings that are both accurate and well ordered.

Perhaps the most consequential finding concerns users who interact rarely with the platform. Sparse data has long been the Achilles’ heel of personalization: users with few recorded actions leave too little evidence for most models to form a reliable interest profile, a phenomenon related to the cold-start and data-scarcity problems documented across the recommender systems literature. On inactive user subsets of the Amazon dataset, the method achieved an NDCG@10 of 0.398, showing that recommendation performance is retained even under the evaluated sparse interaction conditions. The authors attribute this resilience to the fusion design itself, in which weak evidence from one behavior source can be reinforced by complementary evidence from another, so that even a short click history can be enriched by consistent browsing patterns.

The significance of this work sits within a broader research wave on multi-behavior recommendation, where graph neural networks, attention mechanisms, contrastive learning and transformer architectures have all been applied to model interactions among behavior types. Recent studies have explored preference differences among behaviors, cross-attentive behavior-aware graph convolutions, hypergraph-enhanced multi-interest learning and temporal graph transformers, reflecting a consensus that purchase-level feedback alone is too sparse to support high-quality personalization at scale. Zhu’s contribution aligns with this consensus but places particular emphasis on the structural consistency of the shared representation space and the explicit use of temporal semantic constraints, two aspects the author identifies as the limiting factors when multi-source data is modeled insufficiently.

The author is careful to scope the claims. The experiments demonstrate effectiveness within the adopted evaluation setting on the Amazon multi-behavior dataset, and the study notes that the applicability of the learned representations to other e-commerce platforms and different behavioral distributions requires further empirical validation. The paper also reports that no datasets were generated or analyzed during the study beyond those used in the evaluation, and the declared funding for the work is listed as not applicable. Nevertheless, the reported results on inactive users suggest a practical direction for an industry problem that costs platforms real revenue: most visitors to a large online store interact only lightly, and any method that extracts reliable signals from sparse behavioral traces has immediate commercial value.

For the field of ambient intelligence and humanized computing, the study adds to a growing body of evidence that the future of personalization lies not in harvesting ever more data, but in modeling the relationships among the data already collected. As machine learning continues to transform e-commerce, from purchase-intention prediction to sentiment-enhanced recommendation, frameworks that respect the heterogeneity, interdependence and temporal structure of human behavior may prove to be the ones that finally deliver recommendations that feel genuinely personal. The open question, which the study itself flags, is whether interest embeddings learned on one platform’s behavioral distribution will transfer cleanly to another, a challenge that will shape the next generation of multi-source fusion research.

Subject of Research: Personalized e-commerce recommendation using multi-source user behavior data fusion

Article Title: Personalized recommendation methods for e-commerce based on multi-source user behavior data fusion

Article References: Personalized recommendation methods for e-commerce based on multi-source user behavior data fusion. (n.d.). https://doi.org/10.1007/s12652-026-05129-9

Image Credits: AI Generated

DOI: 10.1007/s12652-026-05129-9

Keywords: e-commerce, recommendation systems, multi-behavior recommendation, data fusion, user behavior modeling, temporal dynamics, user interest embeddings, NDCG, sparse data, Amazon dataset, machine learning, personalization

The team employed a hybrid bibliometric–systematic literature review methodology, guided by the PRISMA framework, combining quantitative science mapping with qualitative thematic synthesis. Starting from 162 records retrieved from Scopus on January 28, 2026, two independent reviewers screened titles and abstracts against predefined inclusion and exclusion criteria, ultimately retaining 135 publications across 52 academic sources. The field exhibits an annual growth rate of 14.35 percent, a strikingly young average document age of 2.49 years, and an average of 4.25 citations per document drawn from a cumulative base of 5,946 cited works. Authorship analysis revealed 521 contributing authors with an average of 4.62 co-authors per paper, reflecting the deeply interdisciplinary character of a research area that straddles computer science, marketing, and information systems.

The temporal picture is one of explosive acceleration. Before 2020, scholarly output on AI recommenders in marketplace contexts was sporadic, characteristic of an exploratory phase. A notable increase emerged in 2020, followed by a sharp surge from 2023 onward, with publication peaks of 44 documents in 2024 and 48 in 2025. The authors attribute this trajectory to structural shifts in consumer behavior during and after the COVID-19 pandemic, which accelerated digital adoption and pushed firms to prioritize scalable, automated personalization. As online platforms absorbed enormous volumes of behavioral data, machine learning and deep learning architectures became the default machinery for modeling user–item interactions, transforming marketplaces from transaction-oriented platforms into intelligence-driven ecosystems in which product discovery itself is algorithmically mediated.

Methodologically, the reviewed literature remains dominated by traditional machine learning approaches, prized for their accessibility and modest computational demands. Collaborative filtering and deep learning methods form the second tier, marking a clear shift toward representation learning and data-driven personalization. Deep models—ranging from session-based neural networks to stacked denoising autoencoders—capture nonlinear preference patterns that classical techniques miss, while hybrid architectures that blend multiple recommendation strategies show improved accuracy and resilience against persistent problems such as data sparsity and the cold-start dilemma. Sentiment analysis and natural language processing, including BERT-based frameworks, are increasingly woven into recommendation pipelines, allowing systems to incorporate the emotional and attitudinal signals embedded in reviews and ratings rather than relying solely on transaction histories.

Keyword co-occurrence mapping in VOSviewer revealed six thematic clusters that the authors interpret through a proposed multi-level framework linking AI architecture, consumer cognition, and marketplace implementation. Clusters one and three concern the technological core: recommendation techniques, natural language processing, and predictive analytics that forecast purchase behavior from classification algorithms, random forests, and recurrent neural networks. Clusters two and four address the human side—how personalization intensity, explanation interfaces, and adaptive content shape satisfaction, trust, and purchase intention. Clusters five and six concern platform environments and system integration: interface design, e-service quality, scalable backend architectures, real-time data pipelines, and the governance frameworks required to keep personalization lawful and reliable at industrial scale.

The behavioral analysis draws heavily on two theoretical pillars. The Stimulus–Organism–Response model treats algorithmic features—personalization depth, adaptive ranking, transparency cues—as external stimuli that shape internal cognitive and affective states, which in turn drive engagement and purchasing. The Theory of Planned Behavior explains how attitudes, subjective norms, and perceived behavioral control convert those internal states into intentions. Within this lens, explainable AI emerges as more than a compliance feature: studies show that attribute-based explanations raise user trust and lower algorithmic anxiety in utilitarian shopping contexts, while perceived fairness and privacy protection feed a multidimensional trust construct spanning the recommender itself, the platform, and the individual recommendations it delivers.

Citation analysis exposes the field’s intellectual DNA and its blind spots. The most cited work, a machine learning recommender built on association rule mining by Loukili and colleagues, exemplifies performance-oriented research that prizes predictive accuracy. The second most cited study advances deep neural collaborative filtering, capturing nonlinear preference structures. The third integrates multitask deep learning to predict buying behavior from affective signals in user-generated content. Together these milestones trace an evolution from rule-based optimization to neural architectures to sentiment-aware personalization—yet the authors note that academic recognition remains concentrated on methodological innovation, while trust sustainability, algorithmic bias, and long-term deployment outcomes are comparatively underexplored in the field’s most influential papers.

Perhaps the review’s most consequential finding is the persistent implementation gap between experimental prototypes and production-ready systems. Many algorithms achieve impressive predictive performance in benchmarks, but far fewer studies address infrastructure scalability, data governance, privacy compliance, interoperability, or integration with enterprise architectures. Federated learning approaches and compliance-aware backend designs point toward architectures that can personalize while respecting regulatory constraints, and API-driven integration with unified data normalization is identified as essential for delivering consistent personalization across channels. Geographically, research output is heavily concentrated in India, China, and the United States, with emerging contributions from Indonesia, Morocco, and Malaysia—a pattern the authors link to the maturity of digital market ecosystems and the dominance of fast-turnaround conference venues, which account for roughly 71 percent of the corpus.

The authors also acknowledge the limits of their synthesis. Reliance on a single database may have excluded relevant work indexed elsewhere; the conference-heavy dataset may overrepresent algorithmic advances relative to behavioral theory; and only English-language publications were included, potentially omitting studies from major e-commerce regions. The proposed multi-level framework is explicitly conceptual rather than statistically validated, intended to organize existing knowledge and guide future inquiry. Even so, the synthesis offers a structured roadmap organized around five directions: deeper theoretical integration between behavioral science and algorithm design, longitudinal studies of effects such as algorithm fatigue and over-personalization, ethical research on explainability and bias mitigation, technological work on context-aware and generative personalization, and managerial attention to deployment feasibility.

The overriding message is deceptively simple: a recommender system is only as effective as the weakest of its three interdependent layers. A model that is accurate but opaque erodes trust; a system that is trusted but unscalable never reaches production; an architecture that is scalable but psychologically tone-deaf fails to convert engagement into loyalty. As digital marketplaces pivot from static recommendation lists toward immersive, generative, and real-time personalization, the review argues that the next generation of AI commerce will be judged not by prediction accuracy alone, but by its capacity to be transparent, trustworthy, and deployable—a reframing with profound implications for the platforms that mediate billions of consumer decisions every day.

Subject of Research: A systematic review of AI-based recommender systems in online marketplaces, integrating algorithmic architectures, consumer behavior, and personalization

Article Title: Artificial intelligence recommender systems in online marketplaces integrating architectures consumer behavior and personalization

Article References: Artificial intelligence recommender systems in online marketplaces integrating architectures consumer behavior and personalization. (n.d.). https://doi.org/10.1007/s44163-026-02182-3

Image Credits: AI Generated

DOI: 10.1007/s44163-026-02182-3

Keywords: artificial intelligence, recommender systems, online marketplaces, personalization, consumer behavior, e-commerce, machine learning, deep learning, collaborative filtering, explainable AI, bibliometric analysis, digital commerce