AI can now help to set property values. What are the risks involved?
I came across this paper called Enhancing Explainable AI Land Valuations Reporting for Consistency, Objectivity, and Transparency. By Y. Yim and C. Shing (2025). The paper goes into detail on how explainable AI can ethically support the valuation of properties. This is a sensitive topic, since the valuation of properties affects many parties, like banks, buyers, sellers and the cities. There are major efficiency gains to be made by implementing artificial intelligence and machine learning in this sector. However, their integration also raises legal and ethical questions.
The Core problem
Many models act like a black box; this undermines the duty of the valuer to deliver transparent and consistent valuations. The legal system requires the properties to be thoroughly inspected. and the process to be well-documented.
A possible solution: Making AI explain itself
There are 3 pillars to ethically implement these innovations while integrating these technologies. Consistency: The model should provide repeatable results, following the same process. Objectivity: There should be a clear separation between the developers and validators of a model. Transparency: the model, data and limits should be well documented and easy to understand.
XAI tools like SHAP can be used to make the AI explain itself. These tools show how each feature/variable pushes the price up or down. The chart shown below ranks drivers such as zoning, building age, and floor area. This turns a score into a story that a client can follow.
“The visualisation sets a foundational baseline value (E[f(X)]) of 13.98 on the x-axis, representing the model’s average prediction when no specific feature information is provided. This is the expected value if we were to make a prediction without any additional information. The Output Value (f(x)), which in this instance is 14.07, reflects the actual prediction after accounting for the cumulative effect of the individual features. The colour-coded bars represent the push and pull of each feature on the prediction: the red bars show the features that contribute to an increase in the predicted land value, while the blue bars indicate a decrease.“
Takeaway
AI can increase efficiency and scale valuations. But it must earn trust. Build on three pillars: consistency, objectivity, and transparency. Use SHAP for explanations. Ship reports with a clear checklist and keep human judgement in charge.
References:
Yiu CY, Cheung KS. Enhancing explainable AI land valuations reporting for consistency, objectivity, and transparency. Land. 2025;14(5):927. https://www.proquest.com/scholarly-journals/enhancing-explainable-ai-land-valuations/docview/3212060417/se-2. doi: https://doi.org/10.3390/land14050927.
