Small language models predict user behavior to improve device assistance
Evidence-Grounded SLM Behavioral Prediction Architecture for Future Assistance Use
Artificial Intelligence
Summary
People often change how they interact with devices before asking for help, and predicting this can let a system prepare assistance earlier. The authors found that big language models are too slow and memory-heavy for this task on personal devices. They created a smaller, faster model that looks carefully at user actions in a structured way and checks its own confidence before making predictions. This approach made the models better at guessing when help will be needed while keeping user data private.
What this means in practice
- •For mobile device developers: Provide early and reliable assistance by predicting user help needs directly on smartphones with privacy-preserving small language models.$Commercial implications: Enables smartphone makers to offer anticipatory help features without relying on cloud processing, improving responsiveness and data privacy.
- •For industrial control operators: Anticipate and prepare operator support needs in high-stakes environments by detecting behavior changes with efficient local prediction models.
Authors
Adeyemi Adeseye, Aisvarya Adeseye, Hannu Tenhunen, Jouni Isoaho
Abstract
Users often exhibit changes in interaction behavior before explicitly seeking assistance, enabling systems to anticipate future assistance use. This is especially important in high-stakes environments, where delayed assistance can have operational, financial, or safety consequences. However, conventional Large Language Models (LLMs) are often unsuitable for continuous on-device behavioral prediction because of their computational and memory demands. Behavioral patterns are also highly individual and device-specific, motivating local processing that preserves privacy and data sovereignty. Small Language Models (SLMs) offer a promising alternative, but their ability to reason reliably over low-level sequential interaction events remains underexplored. This study introduces Evidence-Grounded Local Behavioral Prediction (ELBP), a four-layer architecture combining deterministic interaction processing, dual-view behavioral representation, SLM-based prediction, and evidence verification with abstention. Three instruction-tuned models of 3B parameters or fewer were evaluated using raw, structured, dual-view, and complete ELBP configurations. The best-performing model, Qwen2.5-3B with ELBP, achieved a full-set F1-score of 0.80 and PR-AUC of 0.84, representing a 15.9\% relative improvement over raw-event prompting and 17.6\% over XGBoost, while full-set ECE decreased by 43.2\% relative to raw-event prompting. With evidence verification and abstention, selective F1 increased to 0.84 at 91\% coverage and evidence support reached 96\%. These results show that structured behavioral representation and evidence-verified selective prediction can improve the reliability of behavioral forecasting in SLMs.