pith. sign in

arxiv: 2510.22933 · v3 · submitted 2025-10-27 · 💻 cs.CY

How Can AI Augment Access to Justice? Public Defenders' Perspectives on AI Adoption

Inyoung Cheong , Patty Liu , Dominik Stammbach , Peter Henderson This is my paper

Pith reviewed 2026-05-18 03:55 UTC · model grok-4.3

classification 💻 cs.CY
keywords public defendersAI adoptionaccess to justiceevidence investigationlegal AIethical constraintscriminal defensequalitative interviews
0
0 comments X

The pith

Public defenders see AI as most useful for reviewing large volumes of digital evidence but least suitable for courtroom work or defense strategy.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

Public defenders handle overwhelming caseloads while representing clients who often lack other resources. Drawing on interviews with 17 such professionals, the paper maps public defense tasks into five pillars and pinpoints where AI can realistically help without violating ethical duties. Evidence investigation stands out as the clearest opportunity because it involves sifting through massive digital records, while courtroom representation and strategic planning are viewed as requiring irreplaceable human judgment. The work also surfaces practical barriers such as cost, confidentiality rules, and the need for human oversight. A reader should care because better-targeted AI tools could free up time for more thorough representation and thereby improve fairness for people who rely on public defense.

Core claim

Through semi-structured interviews with 17 public defense professionals across the United States, the authors develop a five-pillar map of public defense work—evidence investigation, legal research & writing, client communication & support, courtroom representation, and defense strategies. Interviewees consistently ranked evidence investigation, such as reviewing large volumes of digital records, as the area with greatest potential for AI support, assigned more limited roles to legal research and client communication, and viewed AI as least compatible with courtroom representation and defense strategy. Adoption is constrained by costs, office norms, confidentiality risks, and current tool质量,

What carries the argument

The five-pillar task-level map produced by thematic analysis of the semi-structured interviews, which classifies everyday public defense activities to show where AI assistance is feasible and where it conflicts with ethical or relational requirements.

Load-bearing premise

The perspectives of these 17 self-selected public defense professionals accurately reflect the most important tasks, constraints, and priorities that would apply to AI tools in public defense work more broadly.

What would settle it

A larger, randomly sampled survey of public defenders that ranks AI suitability differently—for example, placing greater emphasis on legal research or client communication—would undermine the reported ordering of opportunities and constraints.

read the original abstract

Public defenders are asked to do more with less: representing clients deserving of adequate counsel while facing overwhelming caseloads and scarce resources. Although artificial intelligence (AI) is often promoted as a means of relieving administrative and cognitive burdens, legal AI research rarely engages with the everyday realities of public defense work. Drawing on in-depth, semi-structured interviews with 17 public defense professionals across the United States, we identify work-intensive tasks most amenable to AI assistance and the ethical constraints involved in legal representation. We develop a comprehensive task-level map of public defense work, dividing it into five pillars to clarify where AI can and cannot contribute: evidence investigation, legal research & writing, client communication & support, courtroom representation, and defense strategies. Interviewees consistently identified evidence investigation, such as reviewing large volumes of digital records, as the area with the greatest potential for AI support. AI was viewed as having more limited roles in legal research and client communication, and as least compatible with courtroom representation and defense strategy. We find that AI adoption is constrained by costs, restrictive office norms, confidentiality risks, and unsatisfactory tool quality. Our interviewees emphasize safeguards for responsible use, including mandatory human verification, limits on over-reliance, and the preservation of relational aspects of lawyering. Building on these findings, we outline a research agenda that promotes equitable access to justice by prioritizing open science, building domain-specific datasets and evaluation, and incorporating frontline practitioners' perspectives into system development.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

0 major / 3 minor

Summary. The paper reports results from in-depth semi-structured interviews with 17 public defense professionals across the United States. It maps public-defense work into five pillars (evidence investigation, legal research & writing, client communication & support, courtroom representation, and defense strategies), finds that interviewees view evidence investigation as the task most amenable to AI support and courtroom representation plus defense strategy as least compatible, identifies adoption constraints (cost, office norms, confidentiality, tool quality), and outlines safeguards plus a research agenda emphasizing open science, domain-specific datasets, and practitioner involvement.

Significance. The study supplies rare, grounded practitioner perspectives on AI use in an under-resourced legal setting. The five-pillar task map and the consistent ranking of task amenability offer concrete guidance for tool design that respects ethical and relational aspects of defense work. The explicit limitations discussion and call for open datasets and frontline input strengthen the paper's utility for future empirical work in AI for access to justice.

minor comments (3)
  1. §3 (Methods): the recruitment description states participants were 'self-selected' but does not report how many offices were initially contacted or the response rate; adding this detail would clarify selection bias without altering the descriptive claims.
  2. §4.2 (Findings on constraints): the paragraph on 'unsatisfactory tool quality' cites only two illustrative quotes; a short table or additional coded excerpts would make the prevalence of this theme more transparent to readers.
  3. §5 (Research agenda): the three proposed directions are listed but not prioritized or linked back to specific interview themes; a brief mapping (e.g., 'open datasets address the evidence-investigation pillar') would tighten the connection.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for their positive review and recommendation to accept the manuscript. We appreciate the recognition of the value of practitioner perspectives from public defenders and the utility of the five-pillar task map for guiding future AI tool development in access-to-justice contexts.

Circularity Check

0 steps flagged

No significant circularity identified

full rationale

The paper reports a qualitative thematic analysis of semi-structured interviews with 17 self-selected public defense professionals. Central claims are explicitly descriptive of participant perspectives on task amenability to AI (evidence investigation highest, courtroom representation and defense strategy lowest) and on adoption constraints. The methods section and limitations discussion acknowledge the modest, non-probability sample and focus on surfacing practitioner views rather than statistical generalization. No derivation gaps, circular reasoning, or unsupported quantitative claims appear in the five-pillar mapping or research agenda. All findings rest on direct interview data and coding rather than any self-referential derivation, fitted parameters, or self-citation chains.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claims rest on the validity and representativeness of qualitative interview data rather than mathematical parameters or new postulated entities.

axioms (1)
  • domain assumption The self-reported views of the 17 interviewed public defense professionals accurately reflect the tasks, constraints, and priorities relevant to AI adoption in public defense.
    The study derives its five-pillar map and ranked AI compatibility directly from these interviews without external validation or deployment testing of actual AI systems.

pith-pipeline@v0.9.0 · 5797 in / 1480 out tokens · 66263 ms · 2026-05-18T03:55:45.151339+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

118 extracted references · 118 canonical work pages · 6 internal anchors

  1. [1]

    Powell v. alabama. 287 U.S. 45, 1932

    work page 1932

  2. [2]

    wainwright

    Gideon v. wainwright. 372 U.S. 335, 1963

    work page 1963

  3. [3]

    land of the free, l.p.https://www4.courts.ca.gov/opinions/documents/B331918

    Noland v. land of the free, l.p.https://www4.courts.ca.gov/opinions/documents/B331918. PDF, September 2025. No. B331918, 2d Dist. Div. 3 (Cal. Ct. App.)

    work page 2025

  4. [4]

    Agentic ai: Autonomous intel- ligence for complex goals–a comprehensive survey.IEEe Access, 2025

    Deepak Bhaskar Acharya, Karthigeyan Kuppan, and B Divya. Agentic ai: Autonomous intel- ligence for complex goals–a comprehensive survey.IEEe Access, 2025

    work page 2025

  5. [5]

    Research and scholarly methods: Semi- structured interviews.Journal of the american college of clinical pharmacy, 4(10):1358–1367, 2021

    Omolola A Adeoye-Olatunde and Nicole L Olenik. Research and scholarly methods: Semi- structured interviews.Journal of the american college of clinical pharmacy, 4(10):1358–1367, 2021

    work page 2021

  6. [6]

    Is your lawyer a lemon? incentives and selection in the public provision of criminal defense.Review of Economics and Statistics, 103 (2):294–309, 2021

    Amanda Agan, Matthew Freedman, and Emily Owens. Is your lawyer a lemon? incentives and selection in the public provision of criminal defense.Review of Economics and Statistics, 103 (2):294–309, 2021

    work page 2021

  7. [7]

    Legal ai: A tool to help lawyers optimize their performance

    Allan Almeida, Shubh Shah, Neer Bhavesh Khandor, and Neha Agarwal. Legal ai: A tool to help lawyers optimize their performance. In2024 International Conference on Electrical Electronics and Computing Technologies (ICEECT), volume 1, pages 1–8. IEEE, 2024

    work page 2024

  8. [8]

    Presenting and evaluating qualitative research.American journal of pharma- ceutical education, 74(8):141, 2010

    Claire Anderson. Presenting and evaluating qualitative research.American journal of pharma- ceutical education, 74(8):141, 2010

    work page 2010

  9. [9]

    How much difference does the lawyer make: The effect of defense counsel on murder case outcomes.Yale LJ, 122:154, 2012

    James M Anderson and Paul Heaton. How much difference does the lawyer make: The effect of defense counsel on murder case outcomes.Yale LJ, 122:154, 2012

    work page 2012

  10. [10]

    Natural language processing for the legal domain: A survey of tasks, datasets, models, and challenges.arXiv preprint arXiv:2410.21306, 2024

    Farid Ariai, Joel Mackenzie, and Gianluca Demartini. Natural language processing for the legal domain: A survey of tasks, datasets, models, and challenges.arXiv preprint arXiv:2410.21306, 2024

    work page arXiv 2024

  11. [11]

    Translat- ing legalese: Enhancing public understanding of court opinions with legal summarizers

    Elliott Ash, Aniket Kesari, Suresh Naidu, Lena Song, and Dominik Stammbach. Translat- ing legalese: Enhancing public understanding of court opinions with legal summarizers. In Proceedings of the 2024 Symposium on Computer Science and Law, CSLAW ’24, page 136–157, New York, NY, USA, 2024. Association for Computing Machinery. ISBN 9798400703331. doi: 10.114...

    work page doi:10.1145/3614407.3643700 2024

  12. [12]

    Aba formal opinion 512: The paradigm for generative ai in legal practice

    The American Bar Association. Aba formal opinion 512: The paradigm for generative ai in legal practice. 2025

    work page 2025

  13. [13]

    Qwen2.5-VL Technical Report

    Shuai Bai, Keqin Chen, Xuejing Liu, Jialin Wang, Wenbin Ge, Sibo Song, Kai Dang, Peng Wang, Shijie Wang, Jun Tang, Humen Zhong, Yuanzhi Zhu, Mingkun Yang, Zhaohai Li, Jianqiang Wan, Pengfei Wang, Wei Ding, Zheren Fu, Yiheng Xu, Jiabo Ye, Xi Zhang, Tianbao Xie, Zesen Cheng, Hang Zhang, Zhibo Yang, Haiyang Xu, and Junyang Lin. Qwen2.5-vl technical report. a...

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  14. [14]

    Gideon at 60: A Snapshot of State Public Defense Systems and Paths to System Reform

    Marea Beeman and Claire Buetow. Gideon at 60: A Snapshot of State Public Defense Systems and Paths to System Reform. Technical Report NIJ Report 307325, National Institute of Jus- tice, U.S. Department of Justice, November 2023. URLhttps://www.ojp.gov/pdffiles1/nij/ 307325.pdf. 18 Real-world Legal Retrieval for Public Defenders

    work page 2023

  15. [15]

    An impact assessment of machine learning risk forecasts on parole board deci- sions and recidivism.Journal of Experimental Criminology, 13(2):193–216, 2017

    Richard Berk. An impact assessment of machine learning risk forecasts on parole board deci- sions and recidivism.Journal of Experimental Criminology, 13(2):193–216, 2017

    work page 2017

  16. [16]

    Springer, 2019

    Richard Berk, Drougas Berk, and D Drougas.Machine learning risk assessments in criminal justice settings. Springer, 2019

    work page 2019

  17. [17]

    Evaluating ai for law: Bridging the gap with open-source solutions.arXiv preprint arXiv:2404.12349, 2024

    Rohan Bhambhoria, Samuel Dahan, Jonathan Li, and Xiaodan Zhu. Evaluating ai for law: Bridging the gap with open-source solutions.arXiv preprint arXiv:2404.12349, 2024

    work page arXiv 2024

  18. [18]

    Incorporating domain knowledge for extractive summarization of legal case docu- ments

    Paheli Bhattacharya, Soham Poddar, Koustav Rudra, Kripabandhu Ghosh, and Saptarshi Ghosh. Incorporating domain knowledge for extractive summarization of legal case docu- ments. InProceedings of the eighteenth international conference on artificial intelligence and law, pages 22–31, 2021

    work page 2021

  19. [19]

    Language models are few-shot learners

    Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhari- wal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agar- wal, Ariel Herbert-Voss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel Ziegler, Jeffrey Wu, Clemens Winter, Chris Hesse, Mark Chen, Eric Sigler, Ma- teusz Litwin, Sc...

    work page 1901

  20. [20]

    Adaptive human-llms interaction col- laboration: Reinforcement learning driven vision-language models for medical report gener- ation

    Yiming Cao, Zhen Li, Lizhen Cui, and Chunyan Miao. Adaptive human-llms interaction col- laboration: Reinforcement learning driven vision-language models for medical report gener- ation. InProceedings of the Extended Abstracts of the CHI Conference on Human Factors in Com- puting Systems, CHI EA ’25, New York, NY, USA, 2025. Association for Computing Machi...

    work page doi:10.1145/3706599.3719852 2025

  21. [21]

    Epistemic and emotional harms of generative ai: Towards human-centered first amendment.https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5435335, 2025

    Inyoung Cheong. Epistemic and emotional harms of generative ai: Towards human-centered first amendment.https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5435335, 2025

    work page 2025

  22. [22]

    (a) i am not a lawyer, but...: engaging legal experts towards responsible llm policies for legal advice

    Inyoung Cheong, King Xia, KJ Kevin Feng, Quan Ze Chen, and Amy X Zhang. (a) i am not a lawyer, but...: engaging legal experts towards responsible llm policies for legal advice. In Proceedings of the 2024 ACM Conference on Fairness, Accountability, and Transparency, pages 2454– 2469, 2024

    work page 2024

  23. [23]

    The tragedy of the master: automation, vulnerability, and distance.Ethics and Information Technology, 17(3):219–229, 2015

    Mark Coeckelbergh. The tragedy of the master: automation, vulnerability, and distance.Ethics and Information Technology, 17(3):219–229, 2015

    work page 2015

  24. [24]

    Reducing workload in systematic review preparation using automated citation classification.Journal of the American Medical Informatics Association, 13(2):206–219, 2006

    Aaron M Cohen, William R Hersh, Kim Peterson, and Po-Yin Yen. Reducing workload in systematic review preparation using automated citation classification.Journal of the American Medical Informatics Association, 13(2):206–219, 2006

    work page 2006

  25. [25]

    Building machines that learn and think with people.Nature human behaviour, 8(10):1851–1863, 2024

    Katherine M Collins, Ilia Sucholutsky, Umang Bhatt, Kartik Chandra, Lionel Wong, Mina Lee, Cedegao E Zhang, Tan Zhi-Xuan, Mark Ho, Vikash Mansinghka, et al. Building machines that learn and think with people.Nature human behaviour, 8(10):1851–1863, 2024. 19 Real-world Legal Retrieval for Public Defenders

    work page 2024

  26. [26]

    Seymour, 2023

    2023) Colorado v. Seymour, 2023. 536 P .3d 1260 (Colo.)

    work page 2023

  27. [27]

    Integrating ai: Guidance and policies for prosecutors, January 2025

    National Best Practices Commission. Integrating ai: Guidance and policies for prosecutors, January 2025. URLhttps://pceinc.org/wp-content/uploads/2025/01/ 20250125-Integrating-AI-A-Guide-for-Prosecutors.pdf

    work page 2025

  28. [28]

    Bias and unfair- ness in information retrieval systems: New challenges in the llm era

    Sunhao Dai, Chen Xu, Shicheng Xu, Liang Pang, Zhenhua Dong, and Jun Xu. Bias and unfair- ness in information retrieval systems: New challenges in the llm era. InProceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, pages 6437–6447, 2024

    work page 2024

  29. [29]

    Mind2Web: Towards a Generalist Agent for the Web

    Xiang Deng, Yu Gu, Boyuan Zheng, Shijie Chen, Samuel Stevens, Boshi Wang, Huan Sun, and Yu Su. Mind2web: Towards a generalist agent for the web, 2023. URLhttps://arxiv.org/ abs/2306.06070

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  30. [30]

    David C. Donald. Bias in ai large language models: Risks and remedies. https://www.americanbar.org/groups/gpsolo/resources/magazine/2025-mar-apr/ bias-ai-large-language-models-risks-remedies/, April 2025

    work page 2025

  31. [31]

    Donohue and Eric A

    John J. Donohue and Eric A. Baldwin. The Crisis of America’s Public Defenders. Project Syndicate, October 2024. URLhttps://www.project-syndicate.org/commentary/ new-public-defender-model-could-transform-us-criminal-justice-by-john-donohue-and-eric-a-baldwin-2024-10

    work page 2024

  32. [32]

    Whispers in the Machine: Confidentiality in Agentic Systems

    Jonathan Evertz, Merlin Chlosta, Lea Schönherr, and Thorsten Eisenhofer. Whispers in the machine: Confidentiality in llm-integrated systems.arXiv preprint arXiv:2402.06922, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  33. [33]

    Natural language processing for lawyers and judges.Mich

    Frank Fagan. Natural language processing for lawyers and judges.Mich. L. Rev., 119:1399, 2020

    work page 2020

  34. [34]

    Levels of autonomy for ai agents sébastien a

    Kevin Feng, David McDonald, and Amy Zhang. Levels of autonomy for ai agents sébastien a. krier using midjourney 6.1.Artificial Intelligence, 2025

    work page 2025

  35. [35]

    Case repositories: Towards case-based reasoning for ai alignment.arXiv preprint arXiv:2311.10934, 2023

    KJ Feng, Quan Ze Chen, Inyoung Cheong, King Xia, and Amy X Zhang. Case repositories: Towards case-based reasoning for ai alignment.arXiv preprint arXiv:2311.10934, 2023

    work page arXiv 2023

  36. [36]

    Privacy-preserving techniques in generative ai and large language models: a nar- rative review.Information, 15(11):697, 2024

    Georgios Feretzakis, Konstantinos Papaspyridis, Aris Gkoulalas-Divanis, and Vassilios S Verykios. Privacy-preserving techniques in generative ai and large language models: a nar- rative review.Information, 15(11):697, 2024

    work page 2024

  37. [37]

    Natural language processing in legal tech.Legal Tech and the Future of Civil Justice (David Engstrom ed.) Forthcoming, 2022

    Jens Frankenreiter and Julian Nyarko. Natural language processing in legal tech.Legal Tech and the Future of Civil Justice (David Engstrom ed.) Forthcoming, 2022

    work page 2022

  38. [38]

    Yoko E Fukumura, Julie McLaughlin Gray, Gale M Lucas, Burcin Becerik-Gerber, and Shawn C Roll. Worker perspectives on incorporating artificial intelligence into office workspaces: Im- plications for the future of office work.International Journal of Environmental Research and Public Health, 18(4):1690, 2021

    work page 2021

  39. [39]

    Theoretical sampling

    Barney G Glaser and Anselm L Strauss. Theoretical sampling. InSociological methods, pages 105–114. Routledge, 2017

    work page 2017

  40. [40]

    Legal tech startup investment is riding high, thanks to ai boost.Crunchbase News, February 2026

    Joanna Glasner. Legal tech startup investment is riding high, thanks to ai boost.Crunchbase News, February 2026. 20 Real-world Legal Retrieval for Public Defenders

    work page 2026

  41. [41]

    Kristen Grauman, Andrew Westbury, Eugene Byrne, Zachary Chavis, Antonino Furnari, Ro- hit Girdhar, Jackson Hamburger, Hao Jiang, Miao Liu, Xingyu Liu, Miguel Martin, Tushar Nagarajan, Ilija Radosavovic, Santhosh Kumar Ramakrishnan, Fiona Ryan, Jayant Sharma, Michael Wray, Mengmeng Xu, Eric Zhongcong Xu, Chen Zhao, Siddhant Bansal, Dhruv Batra, Vincent Car...

    work page arXiv 2022

  42. [42]

    Washington among the bottom of states in public defense funding, 2025

    Jaelyn Grisso. Washington among the bottom of states in public defense funding, 2025. URLhttps://www.cascadepbs.org/investigations/2025/03/ washington-among-bottom-states-public-defense-funding/

    work page 2025

  43. [43]

    A Deep Look into Neural Ranking Models for Information Retrieval

    Jiafeng Guo, Yixing Fan, Liang Pang, Liu Yang, Qingyao Ai, Hamed Zamani, Chen Wu, W. Bruce Croft, and Xueqi Cheng. A deep look into neural ranking models for information retrieval, 2019. URLhttps://arxiv.org/abs/1903.06902

    work page internal anchor Pith review Pith/arXiv arXiv 2019

  44. [44]

    Identifying implicit social biases in vision-language models

    Kimia Hamidieh, Haoran Zhang, Walter Gerych, Thomas Hartvigsen, and Marzyeh Ghassemi. Identifying implicit social biases in vision-language models. InProceedings of the AAAI/ACM Conference on AI, Ethics, and Society, volume 7, pages 547–561, 2024

    work page 2024

  45. [45]

    John Wiley & Sons, 2011

    David Harper and Andrew R Thompson.Qualitative research methods in mental health and psy- chotherapy: A guide for students and practitioners. John Wiley & Sons, 2011

    work page 2011

  46. [46]

    Ai law tracker – fake citations in court (ai hallucinations).https://www

    Peter Henderson. Ai law tracker – fake citations in court (ai hallucinations).https://www. polarislab.org/ai-law-tracker.html, 2025

    work page 2025

  47. [47]

    Pile of law: Learning responsible data filtering from the law and a 256gb open- source legal dataset.Advances in Neural Information Processing Systems, 35:29217–29234, 2022

    Peter Henderson, Mark Krass, Lucia Zheng, Neel Guha, Christopher D Manning, Dan Jurafsky, and Daniel Ho. Pile of law: Learning responsible data filtering from the law and a 256gb open- source legal dataset.Advances in Neural Information Processing Systems, 35:29217–29234, 2022

    work page 2022

  48. [48]

    An empirical study of public defender effectiveness: Self-selection by the marginally indigent.Ohio St

    Morris B Hoffman, Paul H Rubin, and Joanna M Shepherd. An empirical study of public defender effectiveness: Self-selection by the marginally indigent.Ohio St. J. Crim. L., 3:223, 2005

    work page 2005

  49. [49]

    A Common Pool of Privacy Problems: Legal and Technical Lessons from a Large-Scale Web-Scraped Machine Learning Dataset

    Rachel Hong, Jevan Hutson, William Agnew, Imaad Huda, Tadayoshi Kohno, and Jamie Mor- genstern. A common pool of privacy problems: Legal and technical lessons from a large-scale web-scraped machine learning dataset.arXiv preprint arXiv:2506.17185, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  50. [50]

    Clerc: A dataset for legal case retrieval and retrieval-augmented analysis generation, 2024

    Abe Bohan Hou, Orion Weller, Guanghui Qin, Eugene Yang, Dawn Lawrie, Nils Holzenberger, Andrew Blair-Stanek, and Benjamin Van Durme. Clerc: A dataset for legal case retrieval and retrieval-augmented analysis generation, 2024. URLhttps://arxiv.org/abs/2406.17186. 21 Real-world Legal Retrieval for Public Defenders

    work page arXiv 2024

  51. [51]

    CLERC: A dataset for U

    Abe Bohan Hou, Orion Weller, Guanghui Qin, Eugene Yang, Dawn Lawrie, Nils Holzenberger, Andrew Blair-Stanek, and Benjamin Van Durme. CLERC: A dataset for U. S. legal case retrieval and retrieval-augmented analysis generation. In Luis Chiruzzo, Alan Ritter, and Lu Wang, editors,Findings of the Association for Computational Linguistics: NAACL 2025, pages 78...

    work page doi:10.18653/v1/2025.findings-naacl.441 2025

  52. [52]

    Analysis of ai use in courts.https://rails.legal/ resources/resource-ai-orders/, 2025

    Responsible AI in Legal Service. Analysis of ai use in courts.https://rails.legal/ resources/resource-ai-orders/, 2025

    work page 2025

  53. [53]

    Qualitative research method-interviewing and observation.Journal of basic and clinical pharmacy, 5(4):87, 2014

    Shazia Jamshed. Qualitative research method-interviewing and observation.Journal of basic and clinical pharmacy, 5(4):87, 2014

    work page 2014

  54. [54]

    (beyond) reasonable doubt: Challenges that public defenders face in scrutinizing ai in court

    Angela Jin and Niloufar Salehi. (beyond) reasonable doubt: Challenges that public defenders face in scrutinizing ai in court. InProceedings of the 2024 CHI Conference on Human Factors in Computing Systems, CHI ’24, New York, NY, USA, 2024. Association for Computing Machin- ery. ISBN 9798400703300. doi: 10.1145/3613904.3641902. URLhttps://doi.org/10.1145/ ...

    work page doi:10.1145/3613904.3641902 2024

  55. [55]

    Ma- chine bias: There’s software used across the country to predict future criminals

    Surya Mattu Julia Angwin, Jeff Larson, Lauren Kirchner, and ProPublica. Ma- chine bias: There’s software used across the country to predict future criminals. and it’s biased against blacks, May 2016. URLhttps://www.propublica.org/article/ machine-bias-risk-assessments-in-criminal-sentencing

    work page 2016

  56. [56]

    Gpt-4 passes the bar exam.Philosophical Transactions of the Royal Society A, 382(2270):20230254, 2024

    Daniel Martin Katz, Michael James Bommarito, Shang Gao, and Pablo Arredondo. Gpt-4 passes the bar exam.Philosophical Transactions of the Royal Society A, 382(2270):20230254, 2024

    work page 2024

  57. [57]

    Law enforcement and legal professionals’ trust in algorithms.Journal of Law & Empirical Analysis, 2(1):77–96, 2025

    Ryan Kennedy, Lydia Tiede, Amanda Austin, and Kenzy Ismael. Law enforcement and legal professionals’ trust in algorithms.Journal of Law & Empirical Analysis, 2(1):77–96, 2025

    work page 2025

  58. [58]

    Klein, Roberta Calderwood, and Donald G

    Gary A. Klein, Roberta Calderwood, and Donald G. MacGregor. Critical decision method for eliciting knowledge.IEEE Trans. Syst. Man Cybern., 19:462–472, 1989. URLhttps://api. semanticscholar.org/CorpusID:28418547

    work page 1989

  59. [59]

    Sage, 1996

    Steinar Kvale.InterViews: an introduction to qualitive research interviewing. Sage, 1996

    work page 1996

  60. [60]

    Caselaw access project.https://case.law/, n.a

    Harvard Library Innovation Lab. Caselaw access project.https://case.law/, n.a

    work page

  61. [61]

    Ai for defense lawyers, 2025

    UC Berkeley Law. Ai for defense lawyers, 2025. URLhttps://www.law.berkeley.edu/ research/criminal-law-and-justice-center/our-work/ai-for-public-defenders/

    work page 2025

  62. [62]

    Artificial intelligence and the legal profession: Becoming the ai-enhanced lawyer.U

    Michael Legg and Felicity Bell. Artificial intelligence and the legal profession: Becoming the ai-enhanced lawyer.U. Tas. L. Rev., 38:34, 2019

    work page 2019

  63. [63]

    Retrieval-augmented generation for knowledge-intensive nlp tasks.Advances in neural information processing systems, 33:9459–9474, 2020

    Patrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich Küttler, Mike Lewis, Wen-tau Yih, Tim Rocktäschel, et al. Retrieval-augmented generation for knowledge-intensive nlp tasks.Advances in neural information processing systems, 33:9459–9474, 2020. 22 Real-world Legal Retrieval for Public Defenders

    work page 2020

  64. [64]

    LexisNexis. Lexisnexis launches lexis+ ai, a generative ai solution with hallucination-free linked legal citations.https://www.lexisnexis.com/community/pressroom/b/news/posts/ lexisnexis-launches-lexis-ai-a-generative-ai-solution-with-hallucination-free-linked-legal-citations, October 2023

    work page 2023

  65. [65]

    Learning to rank for information retrieval.Foundations and Trends® in Information Retrieval, 3(3):225–331, 2009

    Tie-Yan Liu. Learning to rank for information retrieval.Foundations and Trends® in Information Retrieval, 3(3):225–331, 2009. ISSN 1554-0669. doi: 10.1561/1500000016. URLhttp://dx.doi. org/10.1561/1500000016

    work page doi:10.1561/1500000016 2009

  66. [66]

    Mitigating unintended memorization in language models via alternating teaching

    Zhe Liu, Xuedong Zhang, and Fuchun Peng. Mitigating unintended memorization in language models via alternating teaching. InICASSP 2023-2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 1–5. IEEE, 2023

    work page 2023

  67. [67]

    Chief public defender held in contempt after turning down cases, says office can’t afford it, February 2016

    Andy Lyman. Chief public defender held in contempt after turning down cases, says office can’t afford it, February 2016. URLhttps://nmpoliticalreport.com/2016/12/02/ chief-public-defender-held-in-contempt-after-turning-down-cases-says-office-cant-afford-it/

    work page 2016

  68. [68]

    Integrating large language models in mental health practice: a qualitative descriptive study based on ex- pert interviews.Frontiers in Public Health, 12:1475867, 2024

    Yingzhuo Ma, Yi Zeng, Tong Liu, Ruoshan Sun, Mingzhao Xiao, and Jun Wang. Integrating large language models in mental health practice: a qualitative descriptive study based on ex- pert interviews.Frontiers in Public Health, 12:1475867, 2024

    work page 2024

  69. [69]

    Public defenders as gatekeepers of freedom.UCLA L

    Alma Magaña. Public defenders as gatekeepers of freedom.UCLA L. Rev., 70:978, 2023

    work page 2023

  70. [70]

    Hallucination-free? assessing the reliability of leading ai legal research tools

    Varun Magesh, Faiz Surani, Matthew Dahl, Mirac Suzgun, Christopher D Manning, and Daniel E Ho. Hallucination-free? assessing the reliability of leading ai legal research tools. Journal of Empirical Legal Studies, 22(2):216–242, 2025

    work page 2025

  71. [71]

    Vig, J., Gehrmann, S., Belinkov, Y ., Qian, S., Nevo, D., Singer, Y ., and Shieber, S

    Robert Mahari, Dominik Stammbach, Elliott Ash, and Alex Pentland. The law and NLP: Bridg- ing disciplinary disconnects. In Houda Bouamor, Juan Pino, and Kalika Bali, editors,Findings of the Association for Computational Linguistics: EMNLP 2023, pages 3445–3454, Singapore, Decem- ber 2023. Association for Computational Linguistics. doi: 10.18653/v1/2023.fi...

    work page doi:10.18653/v1/2023.findings-emnlp 2023

  72. [72]

    URLhttps://aclanthology.org/2023.findings-emnlp.224/

    work page 2023

  73. [73]

    LePaRD: A large-scale dataset of judicial citations to precedent

    Robert Mahari, Dominik Stammbach, Elliott Ash, and Alex Pentland. LePaRD: A large-scale dataset of judicial citations to precedent. In Lun-Wei Ku, Andre Martins, and Vivek Srikumar, editors,Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 9863–9877, Bangkok, Thailand, August 2024. Assoc...

    work page doi:10.18653/v1/2024.acl-long.532 2024

  74. [74]

    When trust is zero sum: automation’s threat to epistemic agency.Ethics and Information Technology, 27(2):29, 2025

    Emmie Malone, Saleh Afroogh, Jason D’Cruz, and Kush R Varshney. When trust is zero sum: automation’s threat to epistemic agency.Ethics and Information Technology, 27(2):29, 2025

    work page 2025

  75. [75]

    Nora McDonald, Sarita Schoenebeck, and Andrea Forte. Reliability and inter-rater reliability in qualitative research: Norms and guidelines for cscw and hci practice.Proceedings of the ACM on human-computer interaction, 3(CSCW):1–23, 2019

    work page 2019

  76. [76]

    Generative ai legal landscape 2024, 2024

    Ankit Tandon Megan Ma, Aparna Sinha and Jennifer Richards. Generative ai legal landscape 2024, 2024. 23 Real-world Legal Retrieval for Public Defenders

    work page 2024

  77. [77]

    Human-ai teamwork interface design using patterns of interactions.International Journal of Human–Computer Interaction, 41(11):7112–7134, 2025

    Kazuhiko Momose, Rahul Mehta, Josias Moukpe, Troy R Weekes, and Thomas C Eskridge. Human-ai teamwork interface design using patterns of interactions.International Journal of Human–Computer Interaction, 41(11):7112–7134, 2025

    work page 2025

  78. [78]

    Multi-language transfer learning for low-resource legal case summarization.Artificial Intelligence and Law, 32(4):1111– 1139, 2024

    Gianluca Moro, Nicola Piscaglia, Luca Ragazzi, and Paolo Italiani. Multi-language transfer learning for low-resource legal case summarization.Artificial Intelligence and Law, 32(4):1111– 1139, 2024

    work page 2024

  79. [79]

    The medical profession transformed by artificial intelli- gence: Qualitative study.Digital Health, 8:20552076221143903, 2022

    Lina Mosch, Daniel Fürstenau, Jenny Brandt, Jasper Wagnitz, Sophie AI Klopfenstein, Akira- Sebastian Poncette, and Felix Balzer. The medical profession transformed by artificial intelli- gence: Qualitative study.Digital Health, 8:20552076221143903, 2022

    work page 2022

  80. [80]

    Why ai ‘hallucinations’ are worse than ever.Forbes, June 2023

    Conor Murray. Why ai ‘hallucinations’ are worse than ever.Forbes, June 2023

    work page 2023

Showing first 80 references.