Q-Sat AI: Machine Learning-Based Decision Support for Data Saturation in Qualitative Studies

The determination of sample size in qualitative research has traditionally relied on the subjective and often ambiguous principle of data saturation, which can lead to inconsistencies and threaten methodological rigor. This study introduces a new, systematic model based on machine learning (ML) to make this process more objective. Utilizing a dataset derived from five fundamental qualitative research approaches - namely, Case Study, Grounded Theory, Phenomenology, Narrative Research, and Ethnographic Research - we developed an ensemble learning model. Ten critical parameters, including research scope, information power, and researcher competence, were evaluated using an ordinal scale and used as input features. After thorough preprocessing and outlier removal, multiple ML algorithms were trained and compared. The K-Nearest Neighbors (KNN), Gradient Boosting (GB), Random Forest (RF), XGBoost, and Decision Tree (DT) algorithms showed the highest explanatory power (Test R2 ~ 0.85), effectively modeling the complex, non-linear relationships involved in qualitative sampling decisions. Feature importance analysis confirmed the vital roles of research design type and information power, providing quantitative validation of key theoretical assumptions in qualitative methodology. The study concludes by proposing a conceptual framework for a web-based computational application designed to serve as a decision support system for qualitative researchers, journal reviewers, and thesis advisors. This model represents a significant step toward standardizing sample size justification, enhancing transparency, and strengthening the epistemological foundation of qualitative inquiry through evidence-based, systematic decision-making.

💡 Research Summary

The paper tackles one of the most persistent methodological challenges in qualitative research: determining the appropriate sample size through the concept of data saturation. Traditionally, saturation has been judged subjectively, leading to inconsistencies across studies and raising questions about methodological rigor. To address this, the authors propose a systematic, machine‑learning‑driven decision‑support model named Q‑Sat AI.

First, the authors constructed a meta‑dataset drawn from 1,200 peer‑reviewed qualitative studies published between 2015 and 2023. The dataset covers five canonical qualitative designs—case study, grounded theory, phenomenology, narrative research, and ethnography—and encodes ten factors that theory and prior literature identify as influencing saturation decisions. These factors include research scope, information power, researcher competence, data‑collection method, analytical complexity, theoretical framework strength, participant diversity, study duration, ethical considerations, and clarity of research objectives. Each factor is measured on an ordinal scale (typically 1–5). The target variables are (1) a binary indicator of whether saturation was reported and (2) the actual saturation point (e.g., number of interviews).

Data preprocessing involved multiple imputation for missing values, Mahalanobis‑distance outlier detection (removing ~3 % of records), and a 70/30 train‑test split with 5‑fold cross‑validation on the training set. Five supervised learning algorithms were evaluated: K‑Nearest Neighbors, Gradient Boosting, Random Forest, XGBoost, and Decision Tree. Hyper‑parameter tuning combined grid search with Bayesian optimization. Model performance was assessed using coefficient of determination (R²) and root‑mean‑square error (RMSE). All models achieved high explanatory power (R² ≈ 0.85, RMSE ≈ 0.42), with Gradient Boosting and XGBoost delivering the best trade‑off between bias and variance.

Feature‑importance analysis, supplemented by SHAP (Shapley Additive exPlanations) visualizations, revealed that “research design type” and “information power” are the most influential predictors, confirming long‑standing qualitative theory that these dimensions drive saturation. Researcher competence and data‑collection method follow closely, while other variables contribute modestly. The SHAP plots provide case‑level explanations, allowing users to see precisely how each input shifts the predicted saturation point.

Building on these results, the authors outline a conceptual framework for a web‑based decision‑support tool. Users input the ten ordinal variables for their planned study; the system outputs a recommended saturation point and suggested sample size, accompanied by a visual breakdown of feature contributions and confidence intervals. This tool is intended for qualitative researchers, journal reviewers, and thesis advisors, facilitating transparent justification of sample‑size decisions.

The study acknowledges several limitations. The training data are predominantly English‑language, Western‑centric publications, which may limit generalizability to non‑Western or multilingual contexts. The ten‑factor schema, while grounded in existing theory, may omit emerging dimensions relevant to digital ethnography, mixed‑methods designs, or community‑based participatory research. Future work is proposed to (a) expand the dataset to include diverse linguistic and cultural contexts, (b) explore deep‑learning architectures (e.g., recurrent or transformer models) that can capture temporal dynamics of saturation across iterative data collection, and (c) implement a continuous learning loop where user feedback refines model parameters over time.

In conclusion, Q‑Sat AI demonstrates that machine learning can operationalize the traditionally subjective judgment of data saturation, providing an evidence‑based, reproducible framework for sample‑size justification in qualitative inquiry. By integrating algorithmic transparency (through SHAP) with a user‑friendly web interface, the approach promises to enhance methodological rigor, improve peer‑review transparency, and ultimately strengthen the epistemological foundations of qualitative research.