With
each
year,
technology
helps
us
to
work
smarter
in
some
ways
but
not
so
smart
in
other
ways.
One
critical
area
as
generative

AI
becomes
increasingly
our
personal
assistant
is
to
not
outsource
the
upholding
of
fair
housing
laws. 

Why?

If
you
did
not
know,
generative
AI
already
has
instances
of
contributing
to
and
exacerbating
unfairness
like

here
and

here.
Significant
and
lofty
penalties
have
not
yet
started
being
doled
out
so
now
is
as
good
a
time
as
ever
to
course-set
or
even
course-correct
your
team. 

And,
in
case
you
had
a
moment
to
forget,
the

real
estate
industry
is
more

regulated
than
most
(with
numerous
laws
that
protect
various
demographics)
and
is
facing
scrutiny
on
a
myriad
of
fronts.
As
a
friendly
reminder,
depending
on
where
you
are
in
the
U.S.,
protected
classes
may
include:

• Race
• Color
• Sex 
• Familial
  status 
• National
  origin 
• Disability
  (this
  has
  evolved
  to
  “a
  person
  that
  uses
  an
  assistive
  device”)
• Religion 
• Age 
• Ancestry 
• Sexual
  orientation 
• Gender
  identity 
• Marital
  status 
• Military
  status 
• Domestic
  violence
  victims 
• Source
  of
  income 
• Genetic
  information 
• Pregnancy 
• HIV/AIDS 
• Criminal
  record
  history
• And
  others

In
today’s
litigious
climate,
this
is
the
opportune
time
to
wonder,
“Will
artificial
intelligence
(like
AI
chatbots)
get
me
‘canceled’,
blocked,
fined
or
jailed?” 

Not
if
we
remember
these
9
considerations
for
responsible
AI
in
real
estate:

1. How
   does
   this
   app/tool
   integrate
   
   fair
   housing
   (which
   includes
   fair
   lending)
   laws
   at
   the
   federal,
   state,
   and
   local
   levels?
   Fair
   Housing
   DECODER
   Tip:
   I’ve
   noticed
   that
   some
   of
   the
   most
   popular
   chatbots
   and
   other
   generative
   AI
   include
   the
   federal
   “big
   seven”
   (race,
   color,
   sex,
   familial
   status,
   national
   origin,
   disability,
   religion)
   but
   not
   fair
   housing
   laws
   at
   the
   state
   or
   local
   levels.
2. How
   often
   does
   this
   app/tool
   update
   to
   include
   policy
   changes?
   Fair
   Housing
   DECODER
   Tip:
   Developers
   should
   account
   for
   legal
   changes
   at
   least
   monthly
   as
   there
   have
   been
   numerous
   new
   and
   updated
   fair
   housing
   laws
   and
   case
   laws
   within
   just
   the
   last
   twelve
   months
   across
   the
   U.S. 
3. Did
   the
   developer
   consult
   and
   do
   paired
   testing
   (e.g.
   think
   of
   mystery
   shoppers
   of
   various
   protected
   classes)
   with
   a
   local,
   regional
   or
   national
   fair
   housing
   agency?
4. How
   does
   this
   app/tool
   target
   people
   (such
   as
   a
   “marketing
   avatar”)?
   Fair
   Housing
   DECODER
   Tip:
   B-schools
   teach
   us
   to
   have
   a
   “customer
   avatar”,
   which
   is
   basically
   a
   brand’s
   ideal
   client
   to
   target.
   But,
   fair
   housing
   (and
   again
   this
   includes
   fair
   lending)
   means
   our
   ideal
   client
   cannot
   exclude
   protected
   classes.
   The
   key
   word
   here
   is,
   “exclude”.
   Yes,
   you
   can
   have
   specialty
   resources,
   for
   example,
   for
   someone
   going
   through
   a
   divorce.
   Yet,
   we
   are
   never
   excluding
   (turning
   away)
   those
   who
   are
   not.
5. Are
   the
   “targets”
   based
   on
   any
   fair
   housing
   protected
   class
   (whether
   federally,
   locally
   or
   through
   trade
   organizations)?
   Fair
   Housing
   DECODER
   Tip:
   Use
   tools
   that
   allow
   you
   to
   not
   focus
   on
   the
   features
   of
   people
   but
   rather
   on
   the
   features
   of
   properties
   (“a
   home
   great
   for
   a
   family
   of
   5”
   versus
   “home
   with
   five
   spacious
   bedrooms
   to
   use
   any
   way
   you
   want”).
6. How
   does
   this
   app/tool
   treat
   various
   neighborhoods/zip
   codes?
   Fair
   Housing
   DECODER
   Tip:
   Modern-day
   redlining
   cases
   (c.f.
   
   one
   example)
   show
   companies
   not
   providing
   the
   same
   services
   to
   neighboring
   areas.
   This
   is
   a
   no-no!
7. Does
   it
   “steer”
   people
   with
   one
   set
   of
   demographics
   to
   zip
   codes
   that
   it
   does
   not
   steer
   others?
   Fair
   Housing
   DECODER
   Tip:
   Even
   if
   the
   developer
   has
   not
   done
   paired
   testing,
   your
   team
   can
   do
   paired
   testing!
   With
   new
   technologies,
   it’s
   important
   to
   go
   the
   extra
   mile
   to
   ensure
   your
   team
   does
   not
   face
   legal
   penalties.
8. How
   does
   this
   app/tool
   segment
   into
   niches? 
9. Are
   the
   niches
   based
   on
   protected
   classes?
   Fair
   Housing
   DECODER
   Tip:
   There
   are
   “riches
   in
   niches,”
   but
   also
   “faces
   catch
   cases.”
   Niche
   down
   as
   long
   as
   they
   are
   not
   based
   on
   protected
   demographics.

The
seven
pillars
of
responsible
AI
governance
include

compliance,
trust,
transparency,
fairness,
efficiency,
human
touch
and
reinforced
learning,
which
the
above
questions
encapsulate
to
help
you
start
and
frame
an
AI
partnership. 
In
a
litigious
industry,
if
developers
are
not
willing
to
be
transparent
about
any
of
these
areas
(starting
with
the
eight
questions
above),
it
may
be
worth
your
sanity
to
not
be
an
early
adopter
of
a
particular
platform.

This
column
does
not
necessarily
reflect
the
opinion
of
HousingWire’s
editorial
department
and
its
owners.

To
contact
the
editor
responsible
for
this
piece:



[email protected]

Related