Predicting consumer-technology ownership without a diffusion history

Recent research from arXiv demonstrates a novel approach to predicting consumer technology ownership, even for nascent technologies without historical diffusion data. The study found that perceived attributes of a technology, combined with a log-age covariate, can significantly improve prediction accuracy compared to using only years-since-launch. Notably, ratings from frontier language models (Anthropic Claude Opus 4.7 and OpenAI GPT-5.5) were more effective than human ratings in this predictive task, with Opus 4.7 showing the greatest improvement.

Why it matters

This research provides a data-driven method for anticipating market adoption of new technologies, which is critical for strategic planning, resource allocation, and product development cycles. The ability to forecast ownership early on allows for more informed decision-making regarding investment, market entry, and competitive positioning, particularly in rapidly evolving technological landscapes.

Key insights

• Consumer technology ownership prevalence can be predicted by perceived attributes, even without extensive diffusion history.
• A model using four UTAUT2 acceptance attributes and a log-age covariate significantly improves ownership prediction accuracy over a years-since-launch baseline.
• Human ratings of technology attributes, collected from a 2022 US adult survey, reduced mean absolute prediction error by 17%.
• Frontier language models, specifically Anthropic Claude Opus 4.7 and OpenAI GPT-5.5, provided attribute ratings that further enhanced predictive accuracy beyond human ratings.
• Anthropic Claude Opus 4.7 was the most effective in reducing prediction error for consumer technology ownership.

Source

arXiv — Computers and Society — https://arxiv.org/abs/2608.12344