, 0

Why is Data Science important for your future?

The Big Four Part 3 — DS? (Read the earlier blogs here.)

Easy.  AI is on the way.  It’s gonna be OK.

  • UAD 3.6 is here and forces more detail in development and analysis.
  • Appraiser judgment requires a deeper logic with more detail.
  • Algorithms can be automated, but judgment cannot.
  • Appraisal services become more valuable!

DS – Data Science improves on “automated” appraisals, and is essential for AI – Artificial Intelligence.

DS – Data Science brings on the best appraiser judgment by using AI tools.

AI is an extremely broad world, matching the breadth of the world about us.  Here we focus only on how AI, using DS, creates a surer, faster, more useful appraiser product.  How?

It is not a matter of algorithms versus judgment.  DS integrates and enhances your appraiser judgment.

In legacy appraisal, in the olden days, the appraiser “dug up” comps and made adjustments.  Later, data became much easier – so the appraiser “picked comps.”  Four or five comps was usually enough.  Adjustments were based on “experience” and “familiarity” . . .

Today, AI can do some tasks for us.  It can make some things quite easy!  And it can be dangerous:

  • It can make stuff up, even comps.
  • It can carry forward bias.
  • It can hallucinate.

AI usage can be considered a form of AV “model” (Automated Valuation).

Recall that AV applies preselected algorithms, with judgment restricted to the original programmer’s and developer’s concepts.  Some parameters are assumed, taken for granted, or simply ignored.  The user/client only makes the decisions about “highest use,” acceptable subject features, and expected market segment features.

Any use of AI requires a concise model.  It requires proven algorithms.  And it requires a clear divide between assumptions, human judgment, and AI spin.

So, from where does this concise model come?  Who chooses the AV algorithm?  And what is the risk?

Traditional appraisal is difficult for AI to do:  Decide HBU, pick comps, adjust.  In traditional, legacy appraisal, these all require judgment, good judgment, experienced trained judgment.

Artificial Intelligence needs guidance.  For predictive AI, it needs a path.  It needs detail, guardrails, specified outputs, and adaptation to client/user needs and competence.

The new (FannieMae, FreddieMac) UAD “forms” need this as well.  Detail, Description, Data, and Explanation!

But the new forms lack the explanation which the old requested.  All data, few words.

These GSEs (Government Sponsored Enterprises) can run the detailed data, and find risk/value answers.  But the originators – the banks – will get less explanation.  Less ability to make good risk decisions.  Less.

Good for GSEs, bad for lenders!