Artificial Intelligence (Ai) applications need data science!

This Modernizing Appraisal series is about the analysis process, not the reporting revisions –(New URAR, UAD3.6).  We step through:  scope, data, and ”adjustments.”  It sets today’s path to competence, using data science with artificial intelligence.                                        Read the earlier parts of this series here.

We have looked at the culture and heritage of traditional appraisal.  We now move to data and analyses.

Legacy appraisal practices dominate the literature that the Ai robot might find.  It would discover that “picking comps” is vague, and calculating adjustments is mathematically impossible (except for perhaps “market conditions” or geo-location descriptive contrasting).  “Pick comps, make adjustments.”

Pick comps

A naïve prompt to an Ai robot would say “get some comps.”  In its research, it would find a vast, very vast “body of knowledge” and many, many examples of appraisals using three to five “comparables.”  Ai would find some help in the word-descriptions of a comp in The Appraisal of Real Estate (Appraisal Institute) — which say it should be competitive, similar, and “able to be compared.”

Ai does not do well with sweeping platitudes. Much of appraiser education contains words like “the appraiser should be very careful” and be “worthy of belief.”

Ai does better when given clear instructions!  Given a specific algorithm, it will carry it out exactly.  The computer algorithm is instructed by the appraisal model.  The appraiser decides and directs which model (method) is to be used.

The purpose of data science is to build — with clear instructions, — reliable models, using mathematical logic and calculations.  Ai works better with proper prompts.  Vagueness makes mistakes.

Make adjustments

The typical appraisal problem involves many unknowns (“elements of comparison”) – and not enough data points.  “More unknowns than knowns.”  (Mathematicians call this “the curse of dimensionality.”)

Data science explicitly deals with this problem.  It recognizes that what we do is subject to uncertainty – “unsureness.”  Artificial intelligence continuously deals with levels of certainty – choosing more likely paths or solutions over vague ones (like “be careful” or “be believable”).

Ai can be misled!  Unfortunately, we have marketing promises like “adjustments calculated automatically,” and “prove your adjustments.”  These are simply false advertising, in my opinion.  They are no better than the “adjustment list” your trainer gave you 18 years ago.

The principles of the “Science of Data” are straightforward.  These procedures [principles?] provide a clear, defined path.  They clarify thinking and the ability to understand and explain.

Data science focuses on market analysis, not on “comparing comps.”  Data science depends directly on predictive methods, not believability.  Data science enables the measurement of reliability of the result.  Appraisal data science means a superior product to AVMs, BPOs, Evals, Waivers, and other “hybrids.”

Data science (Evidence Based Valuation) EBV^© enables robust use of the parts of the solution:  identification, data, analysis, report.

EBV optimizes accuracy and reliability, not price and speed.  And enabled by Ai, it is faster, more efficient, and helps prevent the next economic real estate meltdown.

Try it, you’ll like it!