Safety In Numbers

Is big data really able to predict child abuse?
by Gordon Campbell

Big changes are underway for this country’s social safety net. In November, a review panel chaired by Paula Rebstock will report back on the structure and operations of the Child, Youth and Family agency. Already, the government has signalled that the outsourcing – ie privatisation – of some of CYF’s functions is under serious consideration. Big Data is being readied to play a leading role in these imminent changes to the structure and scope of welfare delivery.

The challenge, Finance Minister Bill English explained on September in the course of a Treasury guest lecture on ‘social investment’ lies in changing the nature of the government’s welfare interventions, and analysing whether they’re effective. “This will require the development of the data and measurement infrastructure that delivers the feedback loop to support decision makers…We want to take advantage of developments in data technology and analytics that have transformed many service industries across the economy. We are incorporating these tools in Budget 2016, and we expect in future that they be applied across all social services – that is, becoming systemic and systematic.”

Essentially, English was signalling the government’s plan to take the actuarial tools developed in the insurance and finance industry and apply them to welfare spending. Data is being pooled, and algorithms being developed to help identify behaviour patterns likely to impose future costs on welfare. The potential problems ? Well, using the Big Data approach to detect sources of likely welfare dependence in future raises similar concerns to the ones raised by Edward Snowden about using it to detect the sources of likely security threats in future. Namely, the privacy intrusions involved could well do more harm than good to the families being placed under surveillance – whether that be by the state, or by private contractors.

At the heart of the Rebstock review process is the beguiling vision that child abuse and neglect could (maybe, some day) be statistically predicted with accuracy by the use of Big Data. Supposedly, the data sets are now big enough, sufficient variables can now be included, and the patterns of risk factor combinations are almost within reach. Or at least, they seem close enough to make the attempt look plausible. While algorithms will not be replacing CYF case workers anytime soon, there is a genuine danger (see below) that those data algorithms may soon be driving the nature and the direction of the casework.

Among its terms of reference, the Rebstock review is being asked to explore : “ the potential role of data analytics, including predictive risk modelling, to identify children and young people in need of care and protection.” If that raises concerns about privacy rights (and it should) the review has that covered, too. The terms of reference also require Rebstock to identify: ‘Any legislative barriers that prevent the delivery of improved results for children and young people who come into contact with Child, Youth and Family…In his Treasury speech, Bill English also flagged the government’s willingness to deal with any barriers posed by the public’s current privacy rights : “ The problem to solve is finding the balance between protecting identity and accessing the opportunity to do better. That’s an important part of why the government is establishing the Data Futures Partnership, a permanent organisation that will work on these difficult problems. It is also why we are reviewing the Privacy Act.”

It is certainly a giddy vision, like something torn from the pages of a Philip K. Dick story.

But in reality, can the tools of predictive risk modelling (PRM) truly detect children likely to be abused, and trigger an intervention likely to stop it before it happens? Or more modestly, as per Bill English : will Big Data enable the state to invest in powers of intervention that will be targeted {justifiably or otherwise ) at those families whose PRM scores happen to identify them as being most at risk of abuse and neglect ?

As New Zealand works its way through the theoretical, logistical and ethical tangles involved in the use of PRM as a management tool, the business opportunities of this approach to child protection service detection and delivery are exploding, globally. Across the United States, PRM and its facsimiles are being tested and trialled – and New Zealand is being repeatedly cited as the trail-blazing precedent. English said as much in his Treasury speech : “Our data capability is world class. And we lead the world in implementing the liability and investment approach in welfare.” Here’s Bloomberg News on this country’s development of welfare data as a platform for state intervention:

New Zealand has moved the furthest toward applying data to a range of social services. Along with extra support for children and families, New Zealand’s government is reviewing its driver licensing system after data showed young people in rural areas were being criminally convicted for driving without a valid license, landing them in prison at an annual cost of $87,000 per person. “Every sector is looking for that rich case study and that initial example to be able to think through how something like this might work,” says USC’s [Emily] Putnam-Hornstein. “And New Zealand provides the first example.”

Currently, several US city/state authorities are replicating NewZealand’s interest in statistical tools (within a data-pooling context) in order to enhance welfare targeting. Florida for instance, has its own equivalent called Rapid Safety Feedback (RSF) and has exported it to several other states:

Predictive analytics in child protective services means assigningsuspected abuse cases to different risk levels based on characteristics that have been found to be linked with child abuse. These risk levels can automatically revise as administrative data is updated. Administrative data may be as simple as school reports or could delve deeper into other information that the state holds: the parents’ welfare checks, new criminal offenses or changing marital status…Jurisdictions as far away as New Zealand are also navigating new territory in how administrative welfare data could be used to inform how child abuse calls are screened.

Globally, the private sector is awakening to the profit potential of PRM from (a) the software supporting the statistical calculations and (b) from the use of predictive modelling as a cost-cutting tool in the private sector delivery of child protection services.

A battle for better child maltreatment risk prediction is heating up around the United States and the world. The world’s largest privately held software company, the SAS Institute, has turned its attention to child wellbeing. Leading researchers are turning towards predictive analytics as a tool for child protective service….. Texas is sharing data between its Department of Family and Protective Services and the Department of State Health Services…..Both Los Angeles County and Allegheny County in Pennsylvania are investigating ways that predictive analytics can be used to prevent child maltreatment.

At the same time – in New Zealand, Australia and in the US – voices are sounding a cautionary note about :

(a) the accuracy of the model and the risk of it generating ‘false positives’

(b) the danger of finding a set of statistical co-relations among the suspected perpetrators ( eg being poor, being on a benefit for over 12 months, either parent’s prior personal history of abuse, mental problems, low birth weight, domestic abuse, substance misuse, teenage parenthood ) and treating these factors as the cause of the abusive behaviour in question.

(c) the privacy issues involved when confidential welfare data (about people who have committed no crime) is used to justify state intervention in their lives

(d) the creation of a tier of beneficiary families whose rights are treated as secondary to those of other citizens

(e) the risk of further stigmatising communities and racial groups already suffering from the marginalising effects of poverty.

To repeat: any prospective gains from PRM need to be balanced against the tool’s potential to worsen the social harm that it aims to treat. Circular reasoning and stigmatisation are chronic dangers, all along the spectrum from prediction to substantiation. Some of the inherent difficulties in objectively “substantiating “ the predictions generated by the model are discussed here.

As this US overview article also indicates says, SAS statistical software is underpinning many of the risk assessment tools, and the company’s expansion into child protection services is triggering anxious discussions between policymakers, statisticians, and communities about the future of child protective service delivery:

These conversations are only going to get tougher as big tech companies, sensing fresh markets, ramp up competition with SAS, as communities debate state surveillance, as communities of colour learn whether these tools will improve or exacerbate higher child removal rates, and as social worker judgment starts to clash with statistical output.

Some American journalists have recognised that this trend has serious implications for Maori. This July 20, 2015 article in the US Chronicle of Social Change social work journal makes that point explicitly:

But for all the action in the United States, New Zealand is the furthest along, on the cusp of launching a risk-modeling tool nationwide that would be deployed when a call comes into its hotline… Like Allegheny County [ie the city of Pittsburgh and environs] the model that New Zealand is developing relies on data from multiple agencies to make predictions. While embraced at the top levels of government, there have been some fears that native Māori peoples, already disproportionately represented in the child welfare system, will be heavily targeted by a predictive analytics tool.

When did New Zealand begin heading down this path? The numerical scale of this country’s child protection challenge was summarised in the course of this 2015 article by Massey University social work academic Eileen Oak:

The New Zealand National Census (Statistics New Zealand, 2006) identified 270,000 children living in poverty out of a total population of 4.25 million. Similarly, The White Paper for Vulnerable Children, Vol. II (Ministry of Social Development, 2012) identified that, in 2011, there were 58,000 serious child protection referrals to Child, Youth and Family services (CYF) and, of these, 4,766 cases were of neglect, 3,249 of physical abuse and 1,386 of sexual abuse. In addition, CYF was granted 153,800 Care and Protection Orders…

The current government’s response? In 2011, a Green Paper commissioned by then-Social Development Minister Paula Bennett (and reportedly written by consultant Dr Jo Cribb, now CEO of the Ministry of Women’s Affairs) estimated that 15% of New Zealand children, or about 163,000 could be classified as being ’ vulnerable’ at any one time.

However, as the NZ Heraldreported a year later, the focus narrowed significantly in the subsequent White Paper, onto a far smaller group of welfare recipients.

The concentration of data sharing and statistical predictive tools then began in earnest:

The White Paper, written by a team in the Ministry of Social Development [ without consulting outside experts] is much more prescriptive, but for a much smaller group. It proposes: a national “Vulnerable Kids Information System”, cutely shortened to “Viki”, that will be “a mechanism for extracting and combining information on children (and their caregivers) from existing databases” once a child reaches a “threshold of concern”. Professionals across the sectors “will be able to both view information about these children and enter information about them” …..

Viki was envisaged to become a database on some 20,000 – 30,000 families. (It is not connected to CYRAS, which is CYF’s larger database.) Around this time, MSD contracted a team within Auckland University’s Economics department to assess whether actuarial risk analysis could usefully be deployed on welfare data. The team was headed by associate professor Rhema Vaithianathan, and their seminal September 2012 report is available here:

As Bloomberg News has reported, this early work on PRM has created career opportunities in the US for Vaithianathan, notably in Pittsburgh, Pennsylvania :

In February [ 2015] Vaithianathan, now a professor at Auckland University of Technology, was chosen by Pennsylvania’s Allegheny County to develop models for ranking cases in its child welfare division. Her team will have access to data including county birth records, which indicate whether a baby was born prematurely, if Medicaid paid for the birth, or whether the father was listed on the birth certificate.

Officials in Allegheny County, which encompasses Pittsburgh and its suburbs, say they expect to use the models to help child welfare workers prioritize complaint calls to a hotline for alleged child abuse.

In June 2015, Vaithianathan was reportedly a panelist at a Los Angeles forum called to re-assure the city’s local Community Child Welfare Coalition about the use of personal welfare data for welfare targeting. In US media reports about that meeting, Vaithianathan was described as “the lead researcher developing New Zealand’s predictive risk-modeling tool” – although this is probably no longer the case. Arguably, the key recent New Zealand work on the PRM model is this March 2015 article by Moira Wilson, a principal analyst with the Social Development Ministry.

In one important respect, Wilson’s work marks an improvement on what went before. Instead of relying on historical data about beneficiaries, Wilson’s study drew on a bigger sample from the wider population during a very recent time frame. Interestingly, what it showed is that three risk factors recur amid the general population to the same degree as among beneficiaries : namely, having siblings with contact with CYF, the parents’ own history of CYF contact, and long periods on the benefit.

In several other respects though, Wilson shares some of the same problems as the early Vaithiananthan work. After all, the value of predictive modelling work lies in its alleged ability to accurately predict those cases where child abuse will later be substantiated. Circular reasoning can readily creep into this process, and raise the spectre of racial profiling and stigmatisation. To critics of the PRM approach, that’s a problem with the three main risk factors cited above: families already known to CYF are subject to more surveillance and monitoring, therefore children already known to them are likely to have a flag in the system for subsequent babies born to those parents. “This,” as one social work academic interviewed for this article told me, “would partly explain why contact with CYF for older children and parents’ own CYF histories are such strong predictors. “

This isn’t to say, she added, that abuse isn’t occurring among such families – but if abuse is occurring elsewhere and amid other families, it is less likely to be picked up due to less monitoring; plus if older children have been substantiated, that too is likely to add to the risk picture and will lead to the substantiating of other children from the same families. “Therefore CYF contact becomes self-referential. It predicts itself.”

How does PRM work? As mentioned it applies the actuarial tools for assessing risk in the insurance and finance industries to child welfare. In theory, this needn’t be such a bad thing. If relevant data was taken from the entire birth population and if the sets of risk factors emerged from the data and were not imposed on it from the outset, and if its enthusiasts were not being encouraged to treat a farrago of co-relations as causation and if the subsequent ‘substantiation’ of child abuse and neglect were an entirely objective and value-free exercise, then some of the dangers of circular reasoning, racial profiling and stigmatisation might be avoided, or at least reduced. So far though, the work on refining has been encouraged within a system that appears to have a vested ideological interest in its success.

As for the mechanics involved…put baldly, the early formulations deployed some 132 risk factors from historical runs of data on welfare recipients and these were then checked against later ‘substantiations’ of abuse and neglect – with varying levels of predictive power. Crucially, in the Vaithianathan study that first floated PRM’s boat, the predictions that abuse would occur in the next five years was accurate in only 37 % per cent of cases among the top two deciles held to be at greatest risk.

Meaning : in 63% of the two highest risk categories, “false positives” would be generated – and would have resulted in unjustified interventions in the families concerned if such findings had been acted upon. Nor can this failure be simply put down to it being early days for the model. In the 2015 MSD study by Wilson, the successful predictive rate was, if anything, slightly worse : with the 5% of children with the highest risk scoring, 31.6% of them had been substantiated [of subsequent abuse] by age five years – which obviously means that 69% of those predicted by the model had not.

The problem doesn’t merely lie in the inaccuracy of the predictive tool. PRM tends to define the nexus of things that should be/may be/could be the early warning signs of child abuse via a laundry list of contributory factors so large as to be virtually meaningless. It seems to be a common drawback of this field of work. As British social work researcher Eileen Munro noted in this article:

Lists [of possible causal factors] vary in content and differ in length, with most offering little advice on how factors should be weighted. A systematic review on recognition and response to child neglect identified 74 different risk measurement tools across just 63 studies….

To their credit, Vaithianathan and Wilson have both flagged some of the limitations of the PRM approach. Yet controversially, Vaithianathan also claimed in her ground-breaking 2012 report that the PRM concept is ultimately little different to the screening programmes (for breast cancer etc) carried out routinely in the health system :

The use of Predictive Risk Modelling is most advanced in healthcare and it has not, to our knowledge, been used anywhere [in 2012] in stratifying children based on their risk of maltreatment. However, there is no reason why the same principles that have been successfully applied to the healthcare arena cannot be applied to other areas like child maltreatment.

Really ? Does child abuse and neglect truly manifest itself in society in the same objective way that disease does in the human body? Hardly. There are well known social differences in how abuse gets reported, substantiated and managed, even at a regional level.

Emily Keddell, senior lecturer in social work at Otago University has been among those warning against the political, ethical and practical implications of placing an undue reliance on PRM in the management of social casework. Keddell’s critiques of the PRM approach and its politico-ideological framework can be found here and also here.

As a tool to assist casework, Keddell concludes, PRM could conceivably have uses – but not as the determining factor in family interventions. In her view:

Despite the issues of data integrity, the tool may well identify a group of people who require more assistance and support with parenting. However, from what we know about the relationship between child maltreatment and poverty, this extremely ‘high risk’ group can also be considered to be the product of a society riven with inequalities and lack of social protections. To offer an individualised ‘service’ that does not, indeed cannot, address the broader social policy issues is deeply problematic. It assigns a stigmatised status to individuals in a manner that removes all attention from the wider policy landscape, and implicitly creates a narrative that there are certain people ‘out there’, who are fundamentally different from the rest of us, who have some kind of innate tendency to abuse children that, unless identified and prevented, will eventually express itself like a hidden gene waiting to trigger an inevitable disease process.

As she indicates, the risk of pathologising families in this fashion should be of concern to all of us:

No one wants children to be abused, and a focus on prevention is welcome, but ‘othering’ a sector of the population with no attention to the social landscape that contributes to their problems sorely challenges the ‘effective intervention’ side of the argument, while strengthening the stigma downsides.

Although the advocates of PRM do stress that it should only be seen as a tool – and only as a mere adjunct to social casework – it is easy to see how it could come to drive professional practice. In future, will any caseworker be willing to defer an intervention to remove a child if the family has a high PRM score that might be used against the caseworker – and CYF – should anything later went wrong ? In current practice, the pros and cons of intervention are weighed less mechanically, and perhaps this level of drastic action would be more llikely to be seen as counter-productive. PRM could tip that balance toward pre-emptive action and foster a climate in favour of intervention out of all proportion to the model’s accuracy and predictive power.

Ultimately, the approach strikes Keddell as a wrong-headed way to tackle child abuse and neglect:

‘How can we identify the families who will go on to become abusive?’ is fundamentally the wrong question to ask. A better question is ‘How can we more effectively address the risk factors that contribute to abusive behaviour?’ These are well known and exist across the ecological spectrum. If poverty and drug abuse are both highly co-related with child abuse, then why don’t we reduce poverty, and provide more effective and accessible drug treatment services? If previous contact with CYF as a child is associated with future contact as a parent, then let’s provide better therapeutic and support services to care leavers and other children involved in the care system. If poor communities that are highly transient and with high ratios of children are at greater risk, then make a real investment in community development services to create social cohesion and parenting support groups in such communities. If struggling parents are more likely to access secondary prevention services that are attached to universal services, then let’s provide them….

As mentioned, some of these same concerns are noted in several of the articles that promote PRM research. Unfortunately, they also tend to be tacked on mopre as a cautionary afterthought, rather than integrated with the core of the work. Moira Wilson’s paper, for instance, concludes:

Some, but not all, of the children who go on to have recorded substantiations of maltreatment could be identified early using PRMs. PRMs should be considered as a potential complement to, rather than a replacement for, professional judgment. Trials are needed to establish whether risks can be mitigated and PRMs can make a positive contribution to frontline practice, engagement in preventive services, and outcomes for children.

Onwards and upwards, overall. Despite the caveat:

Deciding whether to proceed to trial requires balancing a range of considerations, including ethical and privacy risks and the risk of compounding surveillance bias.

A similarly schizoid tendency runs through how PRM is being handled at the political level. Incredibly, there is no study of (or research planned into) the incidence of child abuse nationwide. Such a wider study – if it involved PRM – would require access to and analysis of birth data across all of New Zealand during an extended time period, and would need to reach well beyond the welfare statistics – since, after all, child abuse and neglect are not solely a feature of the beneficiary poor. Many of the abuse ‘risk factors’ identified to date occur among families within the paid work force forced to juggle multiple jobs to make ends meet. To state the obvious : child abuse and neglect occurs among all racial groups, and within all levels of income.

However, Social Development Minister Anne Tolley seems well aware that a wider approach – equipped with adequate resourcing – would raise a raft of privacy issues and associated risk of political fallout. In July, Tolley scrapped a proposal for an observational study of a group of 60,000 newborns who would have been (a) classified for levels of risk via the PRM tool, and then (b) later re-assessed to explore if those deemed at high risk had indeed actually gone on to experience abuse.

Ethics approval was going to be sought for the study, but it was immediately halted by a furious Ms Tolley, who wrote in the margins of a document outlining the proposal: “not on my watch, these are children not lab rats”.

Instead, the current intention appears to be for the PRM tool to be restricted to historical welfare data, and in the context of a comparative study as to whether PRM – or ordinary casework – proves to have greater predictive power. Tolley however, can hardly claim to have been oblivious to the ethical debate that had been swirling for years around the uses of PRM.

This briefing for instance, was given to her in November 2014.

The briefing summarised the advice on trialling PRM that had been received from the cross-agency “Predictive Modelling Working Group” ( drawn from MSD, MoH, Te Puna Kokiri and the Children’s Action Plan Directorate) whose work-plan apparently began in March 2014, as well as the response to that advice received from the Vulnerable Children’s Board. While the work-plan outlined was endorsed, the VCB had expressed concerns about the ethical risks involved, and the timeframe proposed. Regardless, a PRM trial was put on the rails at that point – so Tolley can hardly claim outraged innocence six months down the track.

On one level, the current backtracking is surprising. One of the main reasons why the PRM fad has got such traction within the MSD is that it fits the government’s ideological rationale for welfare reform like a glove. By and large, the Key government conceive of the welfare system as providing only a residual safety net in which the reasons for deprivation tend to be attributed to the failings of the individual. Therefore, it is hardly surprising that this minimalist state would want to target its relief measures on a temporary basis (and in largely punitive ways) at those subsisting on the social margins.

For decades, this approach has pre-dominated at the expense of a social investment approach based on the provision of universal services. Deprivation is no longer seen to arise from social causes, and – therefore – welfare support tends to be delivered in ways that promote social atomization, not social cohesion.

To state the obvious: identifying and responding to the social causes of poverty and its ills has not been a hallmark of the Key government’s thinking to date. In recent months there has instead been a rolling drumbeat of official reports that – intentionally or otherwise – have furthered this government’s interest in a reduced role for the state in welfare delivery. Exploring the potential privatization of welfare is part of the terms of reference for the Rebstock review :

The core role and purpose of Child, Youth and Family; and opportunities for a stronger focus on this, including through outsourcing some services

Only weeks ago, the report by the Children’s Commissioner accused the CYF state agency of a “dump and run culture of neglect.” Subsequently, the Productivity Commission called for the individualization and privatization of service delivery.

Supposedly, according to the Productivity Commission, such changes would enable beneficiaries to exercise more “choice” [that old neo-liberal chestnut] among the private providers entrusted with delivering welfare support in future. Ironically, the justification offered for these recommendations included this gem :

The report said Government agencies often tightly prescribed what providers could and could not do, making it difficult for them to innovate and tailor services to an individual’s needs.

Ah-huh. Yet surely, those “ tightly prescribed” failings are the direct result of the Key government’s own demands that its funding of welfare assistance must be strictly rationed, only temporary in duration, renewed repeatedly and targeted via a plethora of qualifying criteria. To condemn the rules that go with that approach – and the frustration that it engenders – as an excuse for going further down the same road of targeting seems to be willfully perverse. Not everyone has been convinced by the Commission’s logic :

The Public Service Association [national secretary Richard Wagstaff ]said the Commission was really trying to introduce more competition into the social service sector, not improve the experience of clients…Experience had shown such processes led to lower labour costs, staff investment and a drop in service quality….[Wagstaff] said the current funding levels were not sufficient. “The workforce is very poorly paid, it has a very high turnover and has inadequate investment in skills.” The Commission’s chair Murray Sherwin said the cost or savings of the proposal was not part of its mandate.

Unfortunately, Sherwin’s comment is par for the course in this sort of exercise. New Zealand has a long track record of setting up reviews to make a case for sweeping change – while carefully fencing the reviewers off from considering whether the funding for the existing system has been sufficient for the job it was expected to do. Starve the state delivery system, and the subsequent weakness can provide a rationale for privatization – in health, in education, and now in welfare delivery. For the sake of the children, of course.

In the meantime, it is not as though the only policy choice is between Big Data and predictive modeling on one hand, and doing absolutely nothing about child abuse and neglect on the other. Other approaches – including say, the Early Start Scheme in Christchurch – have produced promising results. Significantly though, the attempts to leverage that success into a national programme reportedly ran into problems caused by the very same sort of fragmentation (ie. into multiple private providers) that the government and its Productivity Commission are now trying to peddle as a solution to the perceived problems at CYF:

The Auckland University economists who developed the [PRM] model assumed that high-risk families might be offered a home visiting programme such as the US Nurse Family Partnership, which roughly halved the rate of child maltreatment from 54 per cent to 29 per cent. But replicating that success in New Zealand is not so simple. A similar programme in Christchurch, Early Start, has had good success rates, but attempts to roll that out nationally through Family Start have had mixed results because the programme was contracted out to numerous local providers with inadequate training and support.

It is not as if the critics concerned about the evident enthusiasm of government for Big Data methodology – and the approach to welfare problems that it epitomises – do not care about the extent of child abuse and neglect in this country. Many are engaged with it on a daily basis. Their criticism is that the PRM approach fails most of the usual criteria for assessing new tools in social service delivery. According to Jesse Russell, research director at the US National Council of Crime and Delinquency, the tools that we use in social research should be tested on four basic grounds : validity, reliability, equitability and utility.

Validity means that a tool should get the answers as close to right as possible. Yet PRM doesn’t get it right with even close to half the families in the top two deciles deemed to be most at risk. Second, it should be able to be done consistently. Yet with PRM, the sample population and results vary considerably, and are consistent only in the disturbingly high failure rates. Third, the tool should explicitly address equity concerns: does this tool embed racial or socioeconomic inequities or does it alleviate them? Answer : with PRM, Maori and beneficiary families are likely to be further stigmatised by any widespread use of this tool, in its current form. Finally, the answers have to be useful in practice. Here again, the evidence is that in its current form, the PRM tool will do more harm than good to the complex, often abrasive tasks of social casework practice.

Unfortunately, the one thing that PRM would do well is provide a mechanistic way to ration out the delivery of scarce child protection services. Put in the hands of the private sector providers that the government has in mind, PRM may well help them to win service contracts (on cost efficiency grounds) and then enable them to ration the extent and the quality of child protection services accordingly. Do we really want to go down that road?

15 Comments on Safety In Numbers

  1. The article gives the impression that it is well researched when it isn’t. The point of comparison us with the finance and insurance industries. It would have been worthwhile for the author to understand how these techniques work in these settings as many of the concerns stated are common and are addressed in one manner or another. Champion challenger approaches while having understood short comings are proven to be very successful optimising complex multidimensional problems, the alternative is anecdote and unfounded conjecture…..

  2. its incredibly creepy and non human. These types of predictive models have failed to perform accurately.
    There is also an inaccurate insurance model tool the ministry of health developed called the “price of Life ” calculator . This means the ministry has pre predicted how much a person’s LIFE is worth (to them) . Select groups of vulnerable patents have a lower price value calculated for their life = which means for them no medical treatment.
    They routinely use inaccurate and unproven insurance algorithms at the hospital to determine whether or NOT they will fund/provide patients with medical treatment.

    In Cuba St WGTN the ratepayer will fund a surveillance spying (meta data)new technology (collection & prediction) system which violates the Privacy Act. They say this private data collection will make Cuba st safer but that is impossible as it is just meta data collection. Imagine when they use it to arrest innocent Maori men who come out of a bar as the computer has predicted that a crime is going to happen. Data is fudged now, so with inaccurate data an automated system of non response is of concern.
    The ratepayer was not informed that the WCC action-ed the long term plans of centralized global govt (agenda 2030)this goal is much more in the way of surveillance and control technology and (non human)programmed automated systems of response.

  3. What the algorithms eliminate is our humanity.
    We are not programmed robots .
    We are not statistics.
    We are not living in the past but in the living present which is unique.
    Every human being is unique, an INDIVIDUAL, but here they would be subjected to other people’s statistics (and punished) .

    Behaviour modification is used for a reason and the so called predictive program ignores the fact people ( which are not statistics ) can change and many people are unpredictable.
    Individuals .

    To predict something one has to know every variable and have the faculty of intuition.
    For experts in one field we call this state of knowing Zen (and it is a non processing non thinking state).
    A super computer cannot know everything as the human beings that program it do not know everything. Their intention is to save money and to profit off people by sending them to their SERCO private prisons to work/fight, the people behind this don’t see people as living beings but as crime stats/data $$.
    Note there is no plan to stop neglect and abuse. Ironic as for many vulnerable, poor and homeless taxpayers in NZ the govt is found to be both neglectful and abusive. Maybe they should change their behaviour and instead of bullying, neglecting and abusing they could be kind, compassionate and wise. That would set an example for the people in place of the role of corrupt sociopathic corporate weasels .

    To stereotype and use other peoples past meta data is ignorant, wrong and limiting to cause negative outcomes. It is a COLD violation of a human being’s inherent rights targeting select groups of people (liked to IBM & national socialists/Nazis).

  4. Why did this sophisticated risk management PRM tools fail to predict the 2008 crisis or at least safeguard financial institutions from its effects?
    Managers relied too heavily on short-sighted PRM models and too lightly on their own expertise and insight. If anything is to be learned from the crisis, it is that nothing substitutes for human judgment in evaluating business risks and setting the right course of action.’
    The complex PRM’s used did not even have to deal with the unpredictable and ever changing variables of human behaviour.
    I think this is just one example of the failures of PRM’s .

  5. No the private data is not being used for the purpose of its collection. I agree completely its “Minority Report” stuff.

    I also want to note how many judgement people (and right wing eugenic nuts) come out of the woodwork on the issue of welfare,(contraception for woman on welfare), poverty and children. As though these opinionated people think they know the people they are talking about.
    They cannot see the dysfunctional culture and how it ( govt and the welfare business)makes people in a cycle feel unloved, unworthy and disconnected from society.
    When people have a negative image of themselves painted by govt and media (aka “NZ welfare beneficiary” ) and hate themselves they tend to take it out on everyone.

  6. Does any one wonder where these words come from:

    Preemptive strike? Going to war with another sovereign nation on the basis you think they have some intention of using WMDs ( not necessarily intent on on using them on you but on some unnamed country nearby.) No real evidence is needed. Just go in for…some
    Collateral damage? 100s and 1000s of innocent people: dead from war crimes
    No fly Zone? Complete and utter take over of another sovereign states air space and bomb the place unmercifully, with 100s and 1000s more dead
    Quantitative Easing? Adjusting Blame from the feudal bank class and other fat cats that cause improper financial ruin on the poor and marginalised in society by extraction through their pay with taxes and wages in the future.
    Regime change? Philosophical pondering of psychopaths who want to steal other nations resources . THis includes scourging and unmericifully hanging their leaders and killing sometimes millions of its citizens.
    Credit default swop? An insurance scam that bet on the demise of its clients going into receivership???
    Ballout? Looking after the top line Mafia
    Going Forward? Forget your side of the story….we are going to do what we want anyway….going forward!

    Since the industrial revolution we have been sold a pup. There was no evidence the industrial revolution made lives better in fact effects of unemployment, pollution, urban crowding, child labor, and other social ills, the modest rise in average income could well have been accompanied by a fall in the standard of living of the working classes.

    What this article does with precision is show the indifferent hand of government using a sinister tool of gathering data on personal lives in the hope of placing a remote bomb on a target group…. A bit like a smart bomb: not very precise more particularly not very smart. Perhaps a drone?…But I digress: Private prisons have colluded with judges in southern states of US to get prison numbers up.
    This will ensure not only will people get to prison faster by harassment and surveillence they will be sent insane by the police state checking in as well. Perhaps just when this group are wanting to let off steam on a few drinks or a puff they will get pounced on….Ah: easy target practice.

  7. Words are still the primary tools of mind programming.
    We are now in a world full of double speak, social media manipulations, false information overloading and data manipulations. Even $cience theory nowdays is a type of of ‘artificial intelligence’. A political or profitable corporate idea is constructed THEN it is supported by a collection of manipulated data.

    The loads of artificial data( media’s memes & political lies) is downloaded like the “black goo” in the predictive movie “Lucy”.
    It creates an artificial intelligence, a construct, a false perception where “Humanitarian Bombing” is seen not for what it is but just perceived for what the words mean, then accepted by the conditioned mind as the truth.
    The war war 3 escalation justification is done by making it about Putin vs usa. Putin’s invasion of Syria, bombing of Syria with Washington is said to be “good” as he is (allegedly) bombing the “bad usa funded terrorists” and “bad” civilians, making “Bad” Palestinian refugees flee and bombing “bad” Syrian infrastructure.

    I digress too the use of ineffectual PRM’s ” the insurance model for dead static unchanging things” to be used on the dynamic live changeable human being behavior would just be used to create more hardship for the poor, feed the bloated techno & spying corporation’s and overfill the corporation’s private jails.

  8. Don’t the Metservice use algorithms to predict weather? That may be why they are more wrong than right predicting weather.

  9. Weather prediction is interesting in that forcasters 1) pretty well know the complex relationships between variables because they have a physical understanding of how temperature, air, water, the earth’s rotations and gravity interact with each other e.g. hot air rises causing cool air to move in and replace it causing wind and 2) all the variables are observable, if they put monitors in place.

    Abuse prediction monitoring is nothing like that. Users don’t know the complex ways variables relate to each other – there is no magical differential equation for the rate of abuse – they just plug in the variables they have to see if they predict well, without any way of knowing if the variables are causal or just “lucky”. And they include the variables in the model in simple ways e.g. linearly when the real relationship may be much more complex.

    On top of that many important variables are probably not collected or are unable to be collected – age at first consumption of alcohol by the father could be an important variable if it happens at a very young age but the younger it is the less likely it is to be known with any certainty or the person be willing to admit it.

    The best these models can do it predict abuse that people have found out about. The abuse that is not found out would be coded as no abuse which biases the results. This most obvious bias is towards the poor because rich people have the resources to try to hide abuse (e.g. pay the victim to be quiet), the poor don’t.

  10. It all seems like a waste of cash to me. We don’t need to research the risk factors as we already know them
    Benefit dependence
    Family history
    However this government won’t do anything about the prime mover of all these as that would mean developing a policy of generating full employment. And that would contradict all of their neo-liberal beliefs.
    In my experience employment is the key to stable lives and routine.

  11. @ Kelly-Ned its “neocon beliefs” not neoliberal there is nothing “liberal” about the govt (except its corporate welfare policies).

    Many employed wealthy people abuse their children( look at the socioeconomic status of the Whitehall paedophile ring and the gainfully employed UN employees who rape children ). The high paid stable employment and socioeconomic status made their abuse of children “routine”. They have the economic and political positions and power to destroy evidence and coverup the child abuse.

    The violence in the neocon’s agenda, the non inclusive policies perpetrated against the poor and vulnerable is abuse.
    Turning people into debt slaves( low wages in meaningless work for the corporation) without their informed consent is also a form of abuse.
    Selectively ignoring the wealthy and upper middle people abusing their children and covering it up (as is the case in inaccurate PRM models) is used as another blow against the poor .
    That is what this PRM is used for it is propaganda -verbal violence against the poor.

  12. @Kelly-Ned. At the top end of town, just the last three on your list would apply. Only the law, if it has the courage, can intervene in the high socio economic society.

Comments are closed.