Harnessing uncertainty: a new approach to real estate investment decision support

This paper investigates the efficiency of AI-uncertainty quantification as a decision support tool to mitigate adverse selection issues for home buyers and real estate investors. The study is grounded in the premise that buyers can use uncertainty estimates from AI models to identify properties with uncertain price predictions, thus enabling them to filter out risky investment prospects and place bids closer to the fundamental value. We employ an automated valuation model trained on a dataset of over 50,000 historical apartment transactions in Oslo, Norway, spanning 2016–2022. Through a financial performance simulation, we evaluate three bidding strategies: random selection, human model evaluation, and AI-uncertainty quantification. Our findings reveal that AI-uncertainty quantification outperforms human evaluation in identifying uncertain investment opportunities, yielding higher potential profit margins and purchase ratios. The results are robust across different sample sizes and models. Explainable AI analysis using SHAP values further reveals the AI model’s ability to identify spatial uncertainty patterns, enhancing its effectiveness in mitigating adverse selection. Our study contributes to the literature on information asymmetry and automated valuation, offering insights into the application of AI to address adverse selection challenges in residential real estate investment.