Identifying child abuse through text mining and machine learning

Amrit, Chintan; Paauw, Tim; Aly, Robin; Lavrič, Miha

doi:10.1016/j.eswa.2017.06.035

Cited by 89 publications

(47 citation statements)

References 41 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The main area for improvement, though, as hinted at by Schwartz et al . () (see also Amrit et al, ), is the quality, rather than the quantity, of the data used to train algorithms. Each of the projects mentioned above has used what Salganik () refers to as ‘readymade’ datasets.…”

Section: Discussionmentioning

confidence: 98%

“…These problems stand in stark contrast to the healthcare data (both structured and unstructured) used by Amrit et al . ().…”

Section: Discussionmentioning

confidence: 99%

“…Amrit et al . () developed a decision support system to support paediatricians and other healthcare workers in Amsterdam to identify children under their care who are at risk of maltreatment. They used healthcare records, which they claim were complete, and both structured data (entered into a specific field, like age, height and weight) and, as they claim uniquely, unstructured data (the notes made by healthcare workers in free‐text fields).…”

Section: Developing and Implementing Dssmentioning

confidence: 99%

“…Amrit et al . () used a complex set of procedures to test different types of algorithms and found that by combining their approaches to both structured and unstructured data, they produced a tool that was just over 90 per cent accurate. It is planned that the final tool will be implemented in the healthcare system towards the end of 2018.…”

Section: Developing and Implementing Dssmentioning

confidence: 99%

See 3 more Smart Citations

Can Predictive Algorithms Assist Decision‐Making in Social Work with Children and Families?

Gillingham

2019

Child Abuse Review

View full text Add to dashboard Cite

Decision support systems (DSS) which incorporate algorithms trained using administrative data have been promoted as the next promising development in initiatives to improve and assist decision‐making in social work with children and families. In this article, research and grey literature about DSS designed, and in some cases implemented, in the USA, the Netherlands, Australia and New Zealand are analysed to assess the extent to which they are currently able to meet this promise. The challenges of developing DSS for social work are identified and ideas are suggested to overcome them. © 2019 John Wiley & Sons, Ltd. Key Practitioner Messages Decision support systems (DSS) are being developed for social work with children and families and they may transform how decisions are made with important consequences for both service users and the profession. Current DSS are mostly not very accurate in their predictions and further work is required in their development and on the data used to train the algorithms. The next stage in the development of DSS will require long‐term commitment by social workers to work in collaboration with data scientists and researchers. It will also be expensive. Social workers need to lead the debate about whether DSS are worth the investment.

show abstract

Section: Discussionmentioning

confidence: 98%

“…These problems stand in stark contrast to the healthcare data (both structured and unstructured) used by Amrit et al . ().…”

Section: Discussionmentioning

confidence: 99%

Section: Developing and Implementing Dssmentioning

confidence: 99%

Section: Developing and Implementing Dssmentioning

confidence: 99%

See 2 more Smart Citations

Can Predictive Algorithms Assist Decision‐Making in Social Work with Children and Families?

Gillingham

2019

Child Abuse Review

View full text Add to dashboard Cite

show abstract

“…These innovations concern digital participation of parents, but also aim to support professionals' practices. In the recent decade, e-health technologies to alter and support professionals' and citizens' behaviour have become rapidly available, and also machine learning techniques [216] have been developed that can support professionals' early detection of child abuse and neglect. These developments provide new and promising opportunities for the future of safeguarding children.…”

Section: Research Agendamentioning

confidence: 99%

Fragile, please handle with care : understanding and supporting professionals' response to suspicions of child abuse and neglect

Konijnendijk¹

View full text Add to dashboard Cite

Understanding and supporting professionals' response to suspicions of child abuse and neglect FRAGILE, PLEASE HANDLE WITH CARE UNDERSTANDING AND SUPPORTING PROFESSIONALS' RESPONSE TO SUSPICIONS OF CHILD ABUSE AND NEGLECTReferences 181 Summary 193Samenvatting 201 Dankwoord (Acknowledgements) 209 Curriculum Vitae 217Publications and other contributions 221 CHAPTER 1General introduction Child maltreatment traumatises millions of children across the globe [4]. A meta-analysis that included articles published between 1980 and 2008 on the prevalence of child abuse and neglect reported the following overall estimated rates for self-report studies that mainly assessed lifetime prevalence rates: 12.7% for sexual abuse, 22.6% for physical abuse, 36.3% for emotional abuse, 16.3% for physical neglect and 18.4% for emotional neglect. For studies using informants that mainly assessed one-year prevalence rates, the estimated prevalence rates were 0,4% for sexual abuse, and 0,3% for both physical and emotional abuse [4]. Child abuse and neglect in the NetherlandsChild abuse and neglect remains a major societal problem in high-income countries, as a considerable number of children continue to be maltreated and exploited [5]. In the Netherlands, the following definition of child abuse and neglect is currently in use, officially documented in the Kindermishandeling, Veilig Thuis since 2015) on child abuse and neglect reports, and self-reports of students in secondary school. In 2016, the nation-wide prevalence study was limited to self-reports of secondary school students. In 2010, the overall estimated one-year prevalence was 3.4% based on the results of the informants study and the official registration of Veilig Thuis [7]. This prevalence rate is 0.4% higher than found in 2005 [8]. In 2003, the yearly report of the Dutch Advice and Reporting Centre [9] showed the organisation was consulted 47,000 times, and investigated 20,000 suspected cases (0.6% of all Dutch children in 2013 [10]). One-year prevalence rates based on self-report study alone remained stable between 2005 and 2010, 9.9% of students between eleven and seventeen years old reported to have experienced child abuse and neglect [7,8]. The self-report study among secondary school students (in one of the first four years of secondary schools) performed in 2016 (6.5%) showed a decline of the one-year prevalence rate compared to 2005 and 2010 [11].In 2010, the most prevalent types of child maltreatment were physical, emotional and educational neglect, i.e. not attending to children's academic needs. Furthermore, the study concluded that almost a quarter of victims underwent sexual and/or physical abuse in 2010. Compared to 2005, emotional neglect (including being a witness of domestic violence) and educational neglect were reported more often, while sexual abuse and physical neglect were reported less often [7]. Causes and consequencesIn the last two decades, research has increased public knowledge about risk factors and protective factors of child abuse and neglect. It b...

show abstract

Using a Machine Learning Tool to Support High‐Stakes Decisions in Child Protection

Vaithianathan¹,

Benavides-Prado²,

Dalton

et al. 2021

AI Magazine

View full text Add to dashboard Cite

Machine learning decision support tools have become popular in a range of social domains including healthcare, criminal justice, and child welfare. But the design of these tools often fails to consider the potentially complex interactions that happen between the tools and humans. This lack of human‐centered design is one reason that so few tools are actually deployed, and even if they are, struggle to achieve impact. In this article we present the example of the Allegheny Family Screening Tool, a machine learning model used since 2016 to support hotline screening of child maltreatment referrals. We describe aspects of human‐centered design that contributed to the successful deployment of this tool, including agency leadership and ownership, transparency by design, ethical oversight, community engagement, and social license. Finally, we identify potential next‐steps to encourage greater integration of human‐centered design into the development and implementation of machine learning decision support tools.

show abstract

Identifying child abuse through text mining and machine learning

Cited by 89 publications

References 41 publications

Can Predictive Algorithms Assist Decision‐Making in Social Work with Children and Families?

Can Predictive Algorithms Assist Decision‐Making in Social Work with Children and Families?

Fragile, please handle with care : understanding and supporting professionals' response to suspicions of child abuse and neglect

Using a Machine Learning Tool to Support High‐Stakes Decisions in Child Protection

Contact Info

Product

Resources

About