This presentation was a lightening talk (15 min) at the #WIDS2017 conference in Sarasota, FL. It explains how we address intimate partner violence by developing data-driven solutions.
2. Who am I?
Susan Scrupski,
Entrepreneur
30-yr Career in
Technology
Have “lived
experience” with
domestic violence
3. What is Big Mountain Data?
• An early stage social impact startup, founded Fall 2014
• We focus on OFFENDERS and their CRIMES
associated with domestic violence
• We take a STEM approach to solving a societal issue
typically addressed with a victim-centric emotional
appeal for charity ($4B spent annually)
• We partner with leading edge technology companies
that align with our vision and goals
4. Domestic Violence
captures a lot of DATA
The CDC reports that 1 out of 4 women have
experienced severe physical violence from an
intimate partner.
That’s ~40 MILLION women in the U.S.
On the flipside of that equation, there are an equal
number of OFFENDERS. Even if we consider one
batterer stands to abuse 1 - x number of women,
the scale of the “problem” still runs into the millions.
We have data on these offenders today. Further,
we can experiment with new models to merge
unstructured data with the structured data already
in law enforcement databases.
5. What can the data tell us?
• Who the most dangerous repeat offenders are and their criminal
history
• When repeat offenders are more likely to commit an act of violence.
• How at-risk a victim is to being re-victimized by her abuser.
• The probability of whether a first-time offender is a good candidate
for behavioral change.
• Where domestic violence occurs (everywhere).
6. What data?
Data that already exists in law
enforcement systems.
● Calls for Service (Computer-
Aided Dispatch, CAD) 911
calls
● Arrest Data (Record
Management Systems)
● Incident Reports
8. Testing our Thesis
Bayes Impact inaugural hackathon: November
15-16, 2014. Five teams tackled the High Point
PD challenge.
Justice League - Location-aware app that can
discover if offenders that need a
preventative visitation is nearby
Hack DV Offenders - Predictive tool that can
determine the likelihood that a person would
engage in severe domestic violence
Will It Blend - Predictive analytics on who is
likely to engage in domestic violence
To Arrest or Not to Arrest - Machine learning tool
to predict when an arrest should be made
All About the Bayes - Identifies the addresses
where domestic violence is most likely to
happen
10. A response to the Obama
Administration's
President’s Task Force
on 21st Century Policing
14 of 74 of the key issues
identified had to do with data and
transparency
Orlando experimented with the
first community “data dive”
exploring domestic violence and
sexual assault data.
PDI Today:
129 Jurisdictions
Orlando Data Dive
12. Portland Police
Used IBM’s SPSS to assess
the risk of recidivism and
brought the most dangerous
offenders to justice.
“We wanted to find a data-
driven repeatable method that
would help us prioritize the
most important cases without
bias.” - Sergeant Greg Stewart
14. #WhyIStayed
#WhyILeft
An analysis of the social media
phenomenon that erupted over
the Ray Rice NFL scandal.
We analyzed 225K rows of data
to ask a simple question:
WHY DID THEY?
We published the results.
Then, we open sourced the data.