The Usability Scaling Law

The Future of Behavioral Intelligence

The landscape of artificial intelligence is undergoing a rapid transformation, driven in part by the concept of "scaling laws". The widely recognized AI scaling law posits that increasing compute power, training data, and reasoning time leads to increasingly capable AI. Drawing a parallel, Jakob Nielsen proposes a potential "Usability Scaling Law," suggesting that training AI on large volumes of user research data may significantly enhance its ability to predict usability problems and inform better design. This scaling could potentially alter the balance between predictive usability methods (like heuristics) and empirical observation (like user testing), possibly reducing the need for empirical studies in the future for common design tasks.

Much of this deep, contextual knowledge resides as tacit understanding within experienced UX professionals or is scattered across countless projects, internal documents, and forgotten studies.

Historically, usability work has navigated the tension between predicting what designs will succeed and observing actual user behavior. While empirical observation has often been the stronger approach, a vast amount of usability knowledge has been generated over decades. Much of this deep, contextual knowledge resides as tacit understanding within experienced UX professionals or is scattered across countless projects, internal documents, and forgotten studies. Nielsen estimates that potentially a million times more tacit usability knowledge exists than has been formally documented. Accumulating, structuring, and using this immense body of knowledge as AI training data could unlock significant advancements in usability prediction.

However, acquiring and preparing the necessary volume of detailed, annotated data – potentially tens of thousands to a hundred thousand hours of recordings – presents a critical bottleneck. Generating data on this scale is challenging for individual teams and is seen as a task for centralized services or large organizations with extensive research operations.

Developing Behaviorally Intelligent AI

This is where Concrete's approach directly addresses a core challenge in the development of powerful, behaviorally intelligent AI. Concrete has accumulated years of proprietary social behavioral research and nearly two decades of aggregated learnings, generalized knowledge and anonymized data. This extensive history, built over 18 years of work in human-centered design, technology, and communication provides a unique and deep historical perspective. Concrete's data goes beyond conventional usage metrics, specifically designed to capture subtle behavioral patterns and contextual understanding – understanding why users act, not just what they do. This rich, high-value dataset is the essential raw material needed to fuel AI models aimed at understanding, predicting, or simulating human behavior.

Concrete is leveraging this unparalleled foundation to develop "Concrete Behavioral Intelligence," an innovative AI-driven Large Behavior Model (LBM). Unlike conventional AI that often relies solely on transactional data, this model incorporates human psychology, cognitive biases, and emotional triggers to forecast behavior more accurately. The LBM is trained on Concrete's 18 years of structured and unstructured behavioral data from user research, observations, and interviews. By synthesizing these insights with cognitive science models and AI-driven analytics, the model creates a comprehensive, predictive framework.

User research will remain vital for generating new training data to keep AI's knowledge current with evolving user expectations and technological landscapes.

Crucially, while AI prediction is expected to grow, Nielsen emphasizes that empirical observation and user research will not vanish. User research will remain vital for generating new training data to keep AI's knowledge current with evolving user expectations and technological landscapes. It is also necessary for exploring novel interfaces or domain-specific cases where prior data is scarce. This ongoing data refresh is critical for continuous improvement. Concrete has an intact qualitative mixed methods research team poised to conduct research as needed. This team can produce new forms of contextual insights and data essential for the continuous learning and refinement required by AI models, especially agentic and physical AI operating in the real world. They provide the continuous stream of human intelligence needed for refining models and improving real-time reasoning.

Overcome Obstacles that Hindering Productivity and Adoption

Concrete will provide actionable intelligence that facilitates improved decision-making, enhanced user engagement, smarter product design, and more effective marketing. The model will offer deeper audience understanding, predictive behavioral analytics, and insights for optimizing user interfaces and communication strategies. By providing practical, evidence-based recommendations based on predictions, Concrete helps companies overcome obstacles hindering productivity and scaling delivery and adoption. Their expertise helps deliver trusted relationships across omnichannel customer experiences by understanding the deep human and emotional dynamics that create trust. Concrete bridges the gap between traditional slow, manual research and AI analytics that lack deep behavioral context.

In the new intelligence economy, unique, accumulated data is currency.

The substantial market opportunity for advanced behavioral intelligence solutions is underpinned by Concrete's unrivaled competitive advantage: their years of proprietary social behavioral insights and 18 years of aggregated learnings and generalized knowledge and high-value data. In the new intelligence economy, unique, accumulated data is currency, and Concrete's behavioral insights are a critical, high-value asset for any company building an AI model.

By leveraging decades of unique behavioral data to build a predictive LBM and utilizing their research team for continuous, contextual insights, Concrete is directly contributing to the potential of the Usability Scaling Law and the broader field of behavioral intelligence. They provide the essential fuel and the ongoing data stream needed to build AI that truly understands and connects with people.

Read more about Usability Scaling from Jakob Nielsen.