- by rbaz, Data Doctor
- 5/30/2017 2:33:01 PM
This is truly a great ambassador for analytics and the good that can derive from its tools. Usually dollars and cents tend to carry the conversation, but when quality of life is the spear of the initiative, one can't help but be moved.
- by SethBreedlove, Data Doctor
- 5/29/2017 7:31:27 PM
As someone who was a former foster child I see both how wonderful a system like this can be and how the software system itself can be abused. Bluntly put, every family is going to have a period of havoc in their lives where they could be labled a child abuser and there is another human at the end anlayzing the data that can make false assumptions. Once the legal system becomes involved and the machine starts turning it can be very difficult to stop and a well adjusted family can be held up to impossible standards by individuals that really don't have any expertise or good judgment.
On the other hand, there are many child abuse cases that are missed or ignored where this could saves lives.
Also, what is missing in the software are the preventitives such as drug treatment or mental health care that are often but not always the primary starters of child abuse.
- by impactnow, Blogger
- 5/17/2017 11:36:48 AM
This is truly an amazing and heartwarming use of Analytics to make the world a better place. We hear so much about analytics in business for driving profit I love hearing about the great things that analytics can achieve for humanity. This is simply awesome--Thanks, SAS.
- by Lyndon_Henry, Blogger
- 5/16/2017 9:59:37 PM
I think this is an incredible application of analytics and can only imagine the layers of data needed to make this work. It makes me wonder how much data they have access to in order to feed this system. ... The information is out there in a lot of cases even if it is simply in a written compliant that lists all the people present when an incident is reported. Being able to make those connections quickly and assess dangers presented is a Herculean task.
SaneIT is on target in highlighting the complexities involved. The variables in such a predictive system are huge, but in my estimation the most daunting aspect is identifying critical data inputs and then establishing some sort of system not just to collect the data on an ongoing basis, but to ensure its reliability.
- by bkbeverly, Data Doctor
- 5/16/2017 5:02:19 PM
@PC - And just to surface the complexities (many layers) and complications (many obstacles in any given layer), suppose for example Rhode Island wants to proactively prevent child abuse in its state. Let's say that they invest in this system and with the help of their data analysts, discover that kids under the oversight of priests in the lower Massachusettes/Rhode Island area are most likely to be abused. Now you have two issues. Rhode Island has no jurisdiction over the Massachusettes kids and no jurisduiction over the church. In fact, the church has its own policing system whereby priests are not legally prosecuted but the families may be compensated. So now you have this cool system that does what it you want it to do, by producing a set of statistical pointers. However without the legal infrastructure, legal skill sets and political will to bring justice to bear, then the total value of the system is brought into question.
By no means am I questioning the viability of the software, but I am suggesting that in identifying and prosecuting matters such as child abuse, the software purchase is the easiest part of the process. It will point you in a direction based on the data inputs, but as I learned with health care fraud and abuse, the stats are just to get you started. We could easily identify claims filed by health care providers for dead people; you simply match the date of death against the visit date. But we learned that when people are improperly trained to file claims or if a provider works in multiple states that have different rules for Medicaid, then those are not patterns of fraud and abuse. But it takes a lawyer to take the stats and spend time and money to discover that nothing was there. Yes - those systems were marketed as detecting fraud and abuse, but you did not discover the limitations and false-positives until after the license was signed.
In effect, as great as the promise and premise of this software may be, you need to have a legal/investigatory infrastructure already in place to get any real value from it. A divining rod may point me to where the gold mine is. But if I don't have excavation tools, a mineral expert, zoning rights, security, and a method of transporting the gold, then all I have is a good divining rod.
- by PredictableChaos, Data Doctor
- 5/16/2017 2:42:49 PM
@bk - The legal obstalces are interesting.
We have similar problems with self-driving cars. Who is liable for the accident if there is no driver?
We also have similar problems with Uber/Lyft. Will governments allow a service that has lower (or just different) qualification requirements for drivers? Seems the markets want this, but the taxi industry is strongly opposed.
Will certain states see the value in this system, and take the lead in clearing the regulatory obstacles? This is what seems to be happening with self-driving cars. States and cities are rolling out hte red-carpets, so that testing of new vehicles can happen in multiple places.
- by bkbeverly, Data Doctor
- 5/16/2017 12:56:31 PM
While the technical merit of this tool is great, there are three challenges that should be considered. One is cost (which I mentioned in another post). The cost to license, install and configure the software. There is also the cost of training the technical staff to maintain it and the customers to use it. If implemented by a private firm, then the cost would be absorbed or passed off to another party. If implemented by a government agency, then the cost would be passed off as tax increases.
The second is the legal aspect. Child abuse laws vary by state; there are only federal guidelines, hence the actual definition has some wiggle room:
Federal legislation provides guidance to States by identifying a minimum set of acts or behaviors that define child abuse and neglect. The Federal Child Abuse Prevention and Treatment Act (CAPTA) (42 U.S.C.A. § 5106g), as amended by the CAPTA Reauthorization Act of 2010, defines child abuse and neglect as, at minimum:
- "Any recent act or failure to act on the part of a parent or caretaker which results in death, serious physical or emotional harm, sexual abuse or exploitation"; or
- "An act or failure to act which presents an imminent risk of serious harm."
This definition of child abuse and neglect refers specifically to parents and other caregivers. A "child" under this definition generally means a person who is younger than age 18 or who is not an emancipated minor.
While CAPTA provides definitions for sexual abuse and the special cases of neglect related to withholding or failing to provide medically indicated treatment, it does not provide specific definitions for other types of maltreatment such as physical abuse, neglect, or emotional abuse. While Federal legislation sets minimum standards for States that accept CAPTA funding, each State provides its own definitions of maltreatment within civil and criminal statutes.
Third, just as in Fraud and Abuse software, this system can only point to statistical signals that suggest further investigation (context - I have done health care fraud and abuse in my career). Hence, the outputs are just the start of determining the probability of child abuse. And since the definition varies by state, the indicators would have to be general and may reveal some random or explainable noise instead of true abuse indicators.
In short, this software is wonderful, but the cost, need for state-specific customization and need for additional human follow up does not make it a push button solution.
- by kq4ym, Data Doctor
- 5/13/2017 12:19:06 PM
Granted that such a "system creates a "plain English" description of the risk and the factors contributing to the risk score," but unless the staff can be trained and actually has a workload that will allow for the extra steps involved in the analysis and acting on the resulting warnings, it's probably not going to be easy to implement for many understaffed agencies.