Risk Assessment and Technology

1
Risk Assessment and Technology

• Understanding what a risk tool is actually
  telling you... will change how you use technology.

Modeling Solutions, LLC Edow_at_wi.rr.com
2
Edward Dow, Ph.DDirector of Research
Prison Intake Special Management Segregation
Juvenile SVP Consultation

• 20 years
• Clinical Settings
• Residential settings
• Community Mental Health
• Research
• Academics

3
Goals

• Types of risk tools
• General risk assessment tool construction
• Interpretation of results
• Allocation of supervision resources
• Policy implications
• New advances
• Q A

4
Audience Participation

• When do you use risk assessment in your system?
• What do you use?
• Why do you use it?
• What does it do for you?
• How does it affect your allocation of tech
  resources?

5
Principles (Andrews Bonta)

• Risk Target high risk offenders
• Needs Target dysfunctions that are related to
  criminogenic cycle
• Responsivity Offenders will get a different
  amount of therapeutic benefit from a treatment

6
Risk Assessment Tools

• General
• Specialized
• Sex
• Violence
• Clinical/Personality

7
Risk Assessment Construction

• Items presumed to be related to risk
• Assign some weight to them
• Create a scoring scheme
• Determine the distribution of scores
• Define outcome
• Associate scores with outcome
• Validate
• Implement

8
Risk Factor items

• Static Variables that do not change
• Dynamic Variables that do change
• Quasi-Dynamic Variables that change in one
  direction
• What happens when you mix static dynamic?

9
Classical Risk Factor Weighting

• On the average model
• Not specific for any given offender
• Ignore interactive effects

10
Classical Scoring Approach

• Risk variable 1..1
• Risk variable 2..1
• Risk variable 3..1
• Risk variable 4.. -1
• Total 3
• What does the score mean?

11
Classical Scoring Approach

• Risk variable 1-1
• Risk variable 2..1
• Risk variable 3..1
• Risk variable 4..1
• Total 3
• Same score mean different way of getting it.
• What does it mean?

12
Assumption

• Same score same risk

13
Score Distribution
14
Outcome Variables

• If there are only two possible outcome groups-
• There is no Middle Risk

15
Real Life
16
Overlay
17
Under the hood
18
Interpretation

• An offender will be either a recidivist or a
  nonrecidivist.
• Your decision to classify an offender as a
  recidivist or nonrecidivist will represent a low
  medium high level of probability of being
  correct.

19
Allocation of Resources

• What is the purpose of the resource?
• Electronic Monitoring
• Straight up EM
• Passive GPS
• Active GPS
• Kiosks
• Staffing
• Other
• Cost Benefit
• Risk to the community
• Dollar value
• Political

20
Allocation Exercise
21
Principles

• Risk Target high risk offenders- that can
  change
• Needs Target dysfunctions that are related to
  criminogenic cycle for the offender
• Responsivity Offenders will get a different
  amount of therapeutic benefit from a treatment

22
Example-based Modeling

• Patterns Outcomes are better

23
Outcome defines the groups

• Compile good examples of your outcome groups

24
Pattern matching

• Dynamic variables
• Static variables
• Integrity of interactions
• Common Risk Factors

Optimized weights What if analysis Risk
simulations empirically derived
25
Accuracy

• National norm distribution outcome data
• Local norm distribution outcome data
• Patterns outcome data
• Local patterns outcome data

26
AL Pretrial
27
NY
28
WI SE
29
WI Central County EM
30
AL Jail Diversion
31
AL Prison Diversion
32
Policy Implication

• Legislative
• Administrative
• Community Safety

33
Questions?