Analysis of the Pennsylvania Additive Classification Tool: Biases and Important Features

• URL: http://arxiv.org/abs/2112.05860v1
• Date: Fri, 10 Dec 2021 23:00:10 GMT
• Title: Analysis of the Pennsylvania Additive Classification Tool: Biases and Important Features
• Authors: Swarup Dhar and Vanessa Massaro and Darakhshan Mir and Nathan C. Ryan
• Abstract summary: The Pennsylvania Additive Classification Tool (PACT) is used to determine the security level for an incarcerated person in the state's prison system. For a newly incarcerated person it is used in their initial classification. An incarcerated person is reclassified annually using a variant of the PACT and this reclassification can be overridden.
• Score: 0.0
• License: http://creativecommons.org/licenses/by-nc-sa/4.0/
• Abstract: The Pennsylvania Additive Classification Tool (PACT) is a carceral algorithm used by the Pennsylvania Department of Corrections in order to determine the security level for an incarcerated person in the state's prison system. For a newly incarcerated person it is used in their initial classification. The initial classification can be overridden both for discretionary and administrative reasons. An incarcerated person is reclassified annually using a variant of the PACT and this reclassification can be overridden, too, and for similar reasons. In this paper, for each of these four processes (the two classifications and their corresponding overrides), we develop several logistic models, both binary and multinomial, to replicate these processes with high accuracy. By examining these models, we both identify which features are most important in the model and quantify and describe biases that exist in the PACT, its overrides, and its use in reclassification. Because the details of how the PACT operates have been redacted from public documents, it is important to know how it works and what disparate impact it might have on different incarcerated people.

Related papers

• Evaluating the Fairness of Discriminative Foundation Models in Computer Vision [51.176061115977774]
  We propose a novel taxonomy for bias evaluation of discriminative foundation models, such as Contrastive Language-Pretraining (CLIP) We then systematically evaluate existing methods for mitigating bias in these models with respect to our taxonomy. Specifically, we evaluate OpenAI's CLIP and OpenCLIP models for key applications, such as zero-shot classification, image retrieval and image captioning.
  arXiv  Detail & Related papers  (2023-10-18T10:32:39Z)
• Same or Different? Diff-Vectors for Authorship Analysis [78.83284164605473]
  In classic'' authorship analysis a feature vector represents a document, the value of a feature represents (an increasing function of) the relative frequency of the feature in the document, and the class label represents the author of the document. Our experiments tackle same-author verification, authorship verification, and closed-set authorship attribution; while DVs are naturally geared for solving the 1st, we also provide two novel methods for solving the 2nd and 3rd.
  arXiv  Detail & Related papers  (2023-01-24T08:48:12Z)
• PatchMix Augmentation to Identify Causal Features in Few-shot Learning [55.64873998196191]
  Few-shot learning aims to transfer knowledge learned from base with sufficient categories labelled data to novel categories with scarce known information. We propose a novel data augmentation strategy dubbed as PatchMix that can break this spurious dependency. We show that such an augmentation mechanism, different from existing ones, is able to identify the causal features.
  arXiv  Detail & Related papers  (2022-11-29T08:41:29Z)
• Analyzing a Carceral Algorithm used by the Pennsylvania Department of Corrections [0.0]
  This paper is focused on the Pennsylvania Additive Classification Tool (PACT) used to classify prisoners' custody levels while they are incarcerated. The algorithm in this case determines the likelihood a person would endure additional disciplinary actions, can complete required programming, and gain experiences that, among other things, are distilled into variables feeding into the parole algorithm.
  arXiv  Detail & Related papers  (2021-12-06T18:47:31Z)
• Uncertainty in Criminal Justice Algorithms: simulation studies of the Pennsylvania Additive Classification Tool [0.0]
  We study the Pennsylvania Additive Classification Tool (PACT) that assigns custody levels to incarcerated individuals. We analyze the PACT in ways that criminal justice algorithms are often analyzed. We propose and carry out some new ways to study such algorithms.
  arXiv  Detail & Related papers  (2021-12-01T06:27:24Z)
• Rating Facts under Coarse-to-fine Regimes [0.533024001730262]
  We collect 24K manually rated statements from PolitiFact. Our task represents a twist from standard classification, due to the various degrees of similarity between classes. After training, class similarity is sensible over the multi-class datasets, especially in the fine-grained one.
  arXiv  Detail & Related papers  (2021-07-13T13:05:11Z)
• Evaluating Large-Vocabulary Object Detectors: The Devil is in the Details [107.2722027807328]
  We find that the default implementation of AP is neither category independent, nor does it directly reward properly calibrated detectors. We show that the default implementation produces a gameable metric, where a simple, nonsensical re-ranking policy can improve AP by a large margin. We benchmark recent advances in large-vocabulary detection and find that many reported gains do not translate to improvements under our new per-class independent evaluation.
  arXiv  Detail & Related papers  (2021-02-01T18:56:02Z)
• Theoretical Insights Into Multiclass Classification: A High-dimensional Asymptotic View [82.80085730891126]
  We provide the first modernally precise analysis of linear multiclass classification. Our analysis reveals that the classification accuracy is highly distribution-dependent. The insights gained may pave the way for a precise understanding of other classification algorithms.
  arXiv  Detail & Related papers  (2020-11-16T05:17:29Z)
• Learning and Evaluating Representations for Deep One-class Classification [59.095144932794646]
  We present a two-stage framework for deep one-class classification. We first learn self-supervised representations from one-class data, and then build one-class classifiers on learned representations. In experiments, we demonstrate state-of-the-art performance on visual domain one-class classification benchmarks.
  arXiv  Detail & Related papers  (2020-11-04T23:33:41Z)
• Data Augmentation and Clustering for Vehicle Make/Model Classification [0.0]
  We present a way to exploit a training data set of vehicles released in different years and captured under different perspectives. Also the efficacy of clustering to enhance the make/model classification is presented. Deeper convolutional neural network based on ResNet architecture has been designed for the training of the vehicle make/model classification.
  arXiv  Detail & Related papers  (2020-09-14T18:24:31Z)
• Exploiting stance hierarchies for cost-sensitive stance detection of Web documents [24.898077978955406]
  stance detection aims at identifying the position (stance) of a document towards a claim. We propose a modular pipeline of cascading binary classifiers, enabling performance tuning on a per step and class basis. We implement our approach through a combination of neural and traditional classification models that highlight the misclassification costs of minority classes.
  arXiv  Detail & Related papers  (2020-07-29T21:40:01Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.