Pith. sign in

REVIEW 2 major objections 2 minor 80 references

Reviewed by Pith at T0; open to challenge.

T0 means a machine referee read the full paper against a public rubric. The mark states how deep the mechanical check went, never who wrote it. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · grok-4.3

Greater London data shows a U-shaped relationship between crime rates and mental health referrals.

2026-06-26 01:10 UTC pith:6N34IDBW

load-bearing objection The paper spots a U-shaped pattern in London borough data linking mental health referrals to crime rates plus four clusters, but the observational setup does not pin down causality. the 2 major comments →

arxiv 2606.26202 v1 pith:6N34IDBW submitted 2026-06-24 stat.AP

Crime reduction through public healthcare: Interpretable machine learning for mental health service impacts in Greater London

Nadine F\"assler , Ben Moews This is my paper
classification stat.AP
keywords crime ratesmental health referralsU-shaped relationshipGreater Londonmachine learningcluster analysispublic healthcaresocioeconomic deprivation
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper uses statistical methods and interpretable machine learning to analyze links between crime, mental health service access via referrals, and deprivation across London boroughs. It finds a persistent positive association overall, but with a U-shaped pattern where lower service levels may prevent crime while higher levels reflect responses to crime exposure. This matters for policy because it points to potential crime reduction through public healthcare interventions rather than solely law enforcement, and identifies different borough types needing tailored approaches.

Core claim

The analysis reveals a persistent positive association between crime rates and mental health referrals as a proxy for service access. This is contrasted with a nuanced U-shaped relationship suggesting preventive effects at lower service levels and demand-driven responses to crime exposure for higher referral rates. Cluster analysis identifies four borough typologies with distinct combinations of crime rates, mental health service access, and deprivation levels.

What carries the argument

Explainable artificial intelligence techniques and cluster analysis applied to street-level crime data, mental health referral information, and socioeconomic metrics.

Load-bearing premise

Mental health referrals accurately measure service access and the associations are not mainly due to unmeasured socioeconomic or spatial confounding factors.

What would settle it

A study that adds controls for additional spatial and socioeconomic variables and finds the positive association or U-shape disappears would falsify the central claim.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Multifaceted policy approaches are needed instead of universal solutions for different borough types.
  • Preventive mental health interventions may reduce crime at lower service access levels.
  • Crime exposure may increase demand for mental health services at higher referral rates.
  • Interpretable ML can uncover spatial patterns essential for evidence-based policies in public healthcare systems.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Similar U-shaped patterns could be tested in other regions with public healthcare to see if the association holds beyond London.
  • Integrating more granular spatial data might help distinguish causation from correlation in the observed relationships.
  • The findings suggest that increasing mental health access in deprived areas could have crime prevention benefits if the preventive part of the U-shape dominates.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

2 major / 2 minor

Summary. The manuscript analyzes street-level crime data, mental health referrals (as proxy for service access), and socioeconomic metrics across Greater London. It reports a persistent positive association between crime rates and referrals, with a U-shaped relationship interpreted as preventive effects at lower service levels versus demand-driven responses at higher rates; it contrasts this with traditional prevention hypotheses, applies explainable AI to identify crime-category patterns, and uses cluster analysis to define four borough typologies combining distinct crime, access, and deprivation profiles.

Significance. If the associations and U-shape survive rigorous confounding controls, the work would provide one of the first UK-specific explorations of mental-health-service access as a potential crime-related policy lever, underscoring the value of multifaceted rather than uniform interventions. The integration of interpretable ML techniques for pattern discovery is a methodological strength that could improve policy translation.

major comments (2)
  1. [Abstract and Methods description] Abstract and Methods description: the central claim of a U-shaped relationship (preventive at low referrals, demand-driven at high) is load-bearing for the nuanced interpretation, yet no details are supplied on the functional form (e.g., quadratic term, spline, or GAM), the statistical significance or confidence interval around the inflection point, or robustness to alternative specifications; without these, it is impossible to determine whether the shape is data-driven or an artifact of modeling choices.
  2. [Methods description] Methods description: socioeconomic metrics and spatial clustering are included, but the text does not mention spatial lag terms, borough fixed effects, instrumental variables, or difference-in-differences exploiting policy variation; residual spatial or socioeconomic confounding (e.g., policing intensity, reporting norms, or finer-grained deprivation) could therefore generate both the positive slope and the U-shape inflection without reflecting causal regimes.
minor comments (2)
  1. The abstract refers to 'explainable artificial intelligence' without naming the specific post-hoc methods (SHAP, LIME, partial dependence plots, etc.) or the base learners, which would aid reproducibility and allow readers to assess whether interpretability tools themselves could induce the reported patterns.
  2. Clarify the exact temporal coverage of the crime and referral datasets and any preprocessing steps for missing or aggregated values, as these choices directly affect the reliability of the cluster analysis and U-shape estimation.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive and detailed comments, which have prompted us to strengthen the methodological transparency and robustness of the manuscript. We address each major comment below and indicate the revisions made.

read point-by-point responses

  1. Referee: [Abstract and Methods description] Abstract and Methods description: the central claim of a U-shaped relationship (preventive at low referrals, demand-driven at high) is load-bearing for the nuanced interpretation, yet no details are supplied on the functional form (e.g., quadratic term, spline, or GAM), the statistical significance or confidence interval around the inflection point, or robustness to alternative specifications; without these, it is impossible to determine whether the shape is data-driven or an artifact of modeling choices.

    Authors: We agree that explicit details on the functional form, inflection-point statistics, and robustness are essential for evaluating the U-shaped relationship. In the revised manuscript we have expanded the Methods and Results sections to describe the exact functional form used, report the statistical significance and 95% confidence interval around the inflection point, and present robustness checks across alternative specifications (quadratic terms and different smoothing approaches). These additions confirm that the reported shape is data-driven. revision: yes

  2. Referee: [Methods description] Methods description: socioeconomic metrics and spatial clustering are included, but the text does not mention spatial lag terms, borough fixed effects, instrumental variables, or difference-in-differences exploiting policy variation; residual spatial or socioeconomic confounding (e.g., policing intensity, reporting norms, or finer-grained deprivation) could therefore generate both the positive slope and the U-shape inflection without reflecting causal regimes.

    Authors: We acknowledge that additional spatial and fixed-effects controls would further address potential confounding. The original analysis already incorporated socioeconomic metrics and spatial clustering, but did not include spatial lag terms or borough fixed effects. In the revision we have added these as robustness specifications; the main positive association and U-shape remain stable. Instrumental-variable and difference-in-differences approaches are not feasible with the available data, as no suitable exogenous policy variation or instruments for referral rates exist. We have therefore clarified the associational nature of the findings and the corresponding limitations in the revised text. revision: partial

Circularity Check

0 steps flagged

No significant circularity; observational ML analysis derives associations from data without self-referential reductions.

full rationale

The paper applies standard statistical modeling, machine learning, and clustering to street-level crime, referral, and deprivation data across London boroughs. The positive association and U-shaped pattern are presented as outputs of regression and explainable AI techniques applied to the observed variables; no equations, fitted parameters, or cluster definitions are shown to be constructed from the target relationships themselves. No self-citation chains or uniqueness theorems are invoked to justify core modeling choices. The derivation chain remains self-contained against external benchmarks (publicly available administrative records) and does not reduce any claimed result to its own inputs by definition or by renaming.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 0 invented entities

Ledger based on abstract description only; full details on any fitted parameters or additional assumptions not available.

free parameters (1)
  • number of clusters = 4
    Identified four borough typologies via cluster analysis, number likely determined from data exploration.
axioms (2)
  • domain assumption Mental health referrals are a suitable proxy for mental health service access
    Explicitly used as proxy for service access in the analysis.
  • domain assumption Spatial patterns in Greater London data can be meaningfully clustered into typologies
    Cluster analysis performed to identify four types.

pith-pipeline@v0.9.1-grok · 5721 in / 1210 out tokens · 26950 ms · 2026-06-26T01:10:03.845868+00:00 · methodology

0 comments
read the original abstract

The relationship between crime, mental health service access, and socioeconomic deprivation in publicly-funded healthcare systems allowing impactful policy interventions offers an alternative lens to crime prevention that remains underexplored. We address this critical gap through an analysis of street-level crime data, mental health referral information, and socioeconomic metrics across Greater London, using both traditional statistical methods and machine learning techniques to identify relevant relationships and spatial patterns to reveal a persistent positive association between crime rates and mental health referrals as a proxy for service access. The prevailing prevention hypothesis is contrasted with a nuanced U-shaped relationship suggesting a contrast between preventive effects at lower service levels and demand-driven responses to crime exposure for higher referral rates. Subsequent analyses, focussing on explainable artificial intelligence, show distinct crime category patterns, with a cluster analysis identifying four borough typologies with distinct combinations of crime rates, mental health service access, and deprivation levels, requiring multifaceted approaches rather than universal solutions. This research provides one of the first comprehensive studies on this topic for the UK's publicly-funded healthcare system and introduces interpretation-oriented approaches to uncover the patterns essential to evidence-based policies.

Figures

Figures reproduced from arXiv: 2606.26202 by Ben Moews, Nadine F\"assler.

Figure 1
Figure 1. Figure 1: Spatial distribution of variables of interest across London boroughs, excluding the City of London, [PITH_FULL_IMAGE:figures/full_fig_p015_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Simple linear regression results for mental health referrals versus crime rate, per 100,000 population, [PITH_FULL_IMAGE:figures/full_fig_p016_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Decision tree for overall crime rate prediction for Greater London data. The figure shows hierarchi [PITH_FULL_IMAGE:figures/full_fig_p018_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Decision tree for violent crime rate prediction for Greater London data. The figure shows hier [PITH_FULL_IMAGE:figures/full_fig_p019_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Decision tree for property crime rate prediction for Greater London data. The figure shows [PITH_FULL_IMAGE:figures/full_fig_p020_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Cluster analysis results for Greater London, using crime rates, mental health referral rates, and [PITH_FULL_IMAGE:figures/full_fig_p021_6.png] view at source ↗

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

80 extracted references

  1. [1]

    Adler, N. E. and Boyce, T. and Chesney, M. A. and Cohen, S. and Folkman, S. and Kahn, R. L. and Syme, S. L. , title =. American Psychologist , year =

  2. [2]

    Allen, R. C. , title =. American Journal of Economics and Sociology , year =

  3. [3]

    and Eaton, W

    Muntaner, C. and Eaton, W. W. and Miech, R. and O'Campo, P. , title =. Epidemiologic Reviews , year =

  4. [4]

    Becker, G. S. , title =. Journal of Political Economy , year =

  5. [5]

    Berg, M. T. and Stewart, E. A. and Schreck, C. J. and Simons, R. L. , title =. Criminology , year =

  6. [6]

    Bucerius, S. M. and Oriola, T. B. and Jones, D. J. , title =. The Police Journal: Theory, Practice and Principles , year =

  7. [7]

    and Barber, R

    Byrne, A. and Barber, R. and Lim, C. H. , title =. Progress in Neurology and Psychiatry , year =

  8. [8]

    and Danagoulian, S

    Chalfin, A. and Danagoulian, S. and Deza, M. , title =. Journal of Health Economics , year =

  9. [9]

    , title =

    Chandola, T. , title =. Health and Place , year =

  10. [10]

    , title =

    Chen, E. , title =. Current Directions in Psychological Science , year =

  11. [11]

    Choe, J. Y. and Teplin, L. A. and Abram, K. M. , title =. Psychiatric Services , year =

  12. [12]

    Clark, D. M. and Canvin, L. and Green, J. and Layard, R. and Pilling, S. and Janecka, M. , title =. The Lancet , year =

  13. [13]

    Cohen, M. A. , title =. 2020 , publisher =

    2020

  14. [14]

    Cornish, D. B. and Clarke, R. V. , title =. Environmental criminology and crime analysis , year =

  15. [15]

    Corrigan, P. W. and Watson, A. C. , title =. Psychiatry Research , year =

  16. [16]

    Cullen, F. T. and Jonson, C. L. , title =. 2016 , publisher =

    2016

  17. [17]

    Cullen, F. T. and Jonson, C. L. and Nagin, D. S. , title =. The Prison Journal , year =

  18. [18]

    and Farnfield, A

    Delgadillo, J. and Farnfield, A. and North, A. , title =. Counselling and Psychotherapy Research , year =

  19. [19]

    and Lu, T

    Deza, M. and Lu, T. and Maclean, J. C. , title =. Health Economics , year =

  20. [20]

    and Maclean, J

    Deza, M. and Maclean, J. C. and Solomon, K. , title =. Journal of Urban Economics , year =

  21. [21]

    Farrington, D. P. , title =. Criminology , year =

  22. [22]

    and Hayes, A

    Fazel, S. and Hayes, A. J. and Bartellas, K. and Clerici, M. and Trestman, R. , title =. The Lancet Psychiatry , year =

  23. [23]

    , title =

    Andresen, Martin A. , title =. The British Journal of Criminology , volume =. 2006 , doi =

    2006

  24. [24]

    and McGuffog, Ingrid and Western, John S

    Murray, Alan T. and McGuffog, Ingrid and Western, John S. and Mullins, Patrick , title =. The British Journal of Criminology , volume =. 2001 , doi =

    2001

  25. [25]

    and Oliveira, Thiago R

    Suss, Joel H. and Oliveira, Thiago R. , title =. The British Journal of Criminology , volume =. 2023 , doi =

    2023

  26. [26]

    and Swain-Campbell, N

    Fergusson, D. and Swain-Campbell, N. and Horwood, J. , title =. Journal of Child Psychology and Psychiatry , year =

  27. [27]

    , title =

    Field, A. , title =. 2018 , publisher =

    2018

  28. [28]

    and Firth, N

    Finegan, M. and Firth, N. and Delgadillo, J. , title =. Psychotherapy Research , year =

  29. [29]

    Fowler, P. J. and Tompsett, C. J. and Braciszewski, J. M. and Jacques-Tiura, A. J. and Baltes, B. B. , title =. Development and Psychopathology , year =

  30. [30]

    and Winters, P

    Franks, P. and Winters, P. C. and Tandredi, D. J. and Fiscella, K. A. , title =. BMC Cardiovascular Disorders , year =

  31. [31]

    Galloway, T. A. and Skardhamar, T. , title =. European Journal of Criminology , year =

  32. [32]

    , title =

    Garland, D. , title =. Critical Review of International Social and Political Philosophy , year =

  33. [33]

    and Tzani-Pepelasi, C

    Halle, C. and Tzani-Pepelasi, C. and Pylarinou, N. R. and Fumagalli, A. , title =. New Ideas in Psychology , year =

  34. [34]

    , title =

    Hannon, L. , title =. Sociological Spectrum , year =

  35. [35]

    Hastie, T. J. and Tibshirani, R. and Friedman, J. H. , title =. 2009 , publisher =

    2009

  36. [36]

    and Stock, W

    Hendrix, L. and Stock, W. A. , title =. The Journal of Human Resources , year =

  37. [37]

    and Schomaker, M

    Heumann, C. and Schomaker, M. and Shalabh , title =. 2016 , publisher =

    2016

  38. [38]

    Hipp, J. R. and Yates, D. K. , title =. Criminology , year =

  39. [39]

    and Mednick, S

    Hodgins, S. and Mednick, S. A. and Brennan, P. A. and Schulsinger, F. and Engberg, M. , title =. Archives of General Psychiatry , year =

  40. [40]

    The British Journal of Criminology , volume =

    Jackson, Jonathan and Stafford, Mai , title =. The British Journal of Criminology , volume =. 2009 , doi =

    2009

  41. [41]

    Hoeboer, C. M. and Kitselaar, W. M. and Henrich, J. F. and Miedzobrodzka, E. J. and Wohlstetter, B. and Giebels, G. and Meynen, G. and Kruisbergen, E. W. and Kempes, M. and Olff, M. and De Kogel, C. H. , title =. American Journal of Criminal Justice , year =

  42. [42]

    2021 , doi =

    Filaments of crime: Informing policing via thresholded ridge estimation , journal =. 2021 , doi =

    2021

  43. [43]

    and Witten, D

    James, G. and Witten, D. and Hastie, T. and Tibshirani, R. , title =. 2021 , publisher =

    2021

  44. [44]

    , title =

    Kelly, M. , title =. The Review of Economics and Statistics , year =

  45. [45]

    and Lee, S

    Kim, S. and Lee, S. , title =. Sustainable Cities and Society , year =

  46. [46]

    and Sembajwe, G

    Kondo, N. and Sembajwe, G. and Kawachi, I. and Van Dam, R. M. and Subramanian, S. V. and Yamagata, Z. , title =. BMJ , year =

  47. [47]

    Kruskal, W. H. and Wallis, W. A. , title =. Journal of the American Statistical Association , year =

  48. [48]

    and Johnson, K

    Kuhn, M. and Johnson, K. , title =. 2013 , publisher =

    2013

  49. [49]

    , title =

    Carr, N. , title =. Probation Journal , volume =. 2017 , doi =

    2017

  50. [50]

    and Ki, D

    Lee, S. and Ki, D. and Hipp, J. R. and Kim, J. H. , title =. Urban Studies , year =

  51. [51]

    Link, B. G. and Andrews, H. and Cullen, F. T. , title =. American Sociological Review , year =

  52. [52]

    Miller, T. R. and Cohen, M. A. and Swedler, D. I. and Ali, B. and Hendrie, D. V. , title =. Journal of Benefit-Cost Analysis , year =

  53. [53]

    Miller, T. R. and Cohen, M. A. and Wiersema, B. , title =. 1996 , number =

    1996

  54. [54]

    Moore, M. H. , title =. Crime and Justice , year =

  55. [55]

    Moran, P. A. , title =. Biometrika , year =

  56. [56]

    and Knowles, C

    Ip Tat Kuen, C. and Knowles, C. and Hardy, C. and Murray, E. , title =. 2017 , number =

    2017

  57. [57]

    , title =

    Newburn, T. , title =. Crime and Justice , year =

  58. [58]

    Peterson, R. D. and Krivo, L. J. , title =. 2010 , publisher =

    2010

  59. [59]

    Pickett, K. E. and Wilkinson, R. G. , title =. The British Journal of Psychiatry , year =

  60. [60]

    The relationship between mental disorders and types of crime in inmates in a

    Pond\'. The relationship between mental disorders and types of crime in inmates in a. Journal of Forensic Sciences , year =

  61. [61]

    , title =

    Pratt, J. , title =. The British Journal of Criminology , year =

  62. [62]

    Pratt, T. C. and Cullen, F. T. , title =. Crime and Justice , year =

  63. [63]

    and Breno, A

    Ramezani, N. and Breno, A. J. and Mackey, B. J. and Viglione, J. and Evans Cuellar, A. and Johnson, J. E. and Taxman, F. S. , title =. BMC Health Services Research , year =

  64. [64]

    , title =

    Rudin, C. , title =. Nature Machine Intelligence , year =

  65. [65]

    and Roccato, M

    Russo, S. and Roccato, M. , title =. Journal of Community Psychology , year =

  66. [66]

    Sampson, R. J. and Wilson, W. J. , title =. Crime and inequality , editor =. 1995 , publisher =

    1995

  67. [67]

    Shapiro, S. S. and Wilk, M. B. , title =. Biometrika , year =

  68. [68]

    , title =

    Sheather, S. , title =. 2009 , publisher =

    2009

  69. [69]

    and Loureiro, A

    Silva, M. and Loureiro, A. and Cardoso, G. , title =. The European Journal of Psychiatry , year =

  70. [70]

    and Felson, R

    Silver, E. and Felson, R. B. and VanEseltine, M. , title =. Criminal Justice and Behavior , year =

  71. [71]

    , title =

    Steinley, D. , title =. Psychological Methods , year =

  72. [72]

    Swanson, J. W. and Holzer III, C. E. and Ganju, V. K. and Jono, R. T. , title =. Psychiatric Services , year =

  73. [73]

    and Piquero, A

    Sweeten, G. and Piquero, A. R. and Steinberg, L. , title =. Journal of Youth and Adolescence , year =

  74. [74]

    Vinkers, D. J. and de Beurs, E. and Barendregt, M. and Rinne, T. and Hoek, H. W. , title =. Criminal Behaviour and Mental Health , year =

  75. [75]

    Wagner, K. L. , title =. The B.E. Journal of Economic Analysis and Policy , year =

  76. [76]

    Walker, J. T. , title =. Justice Quarterly , year =

  77. [77]

    and Hockenberry, J

    Wen, H. and Hockenberry, J. M. and Cummings, J. R. , title =. Journal of Public Economics , year =

  78. [78]

    Wilkinson, R. G. and Pickett, K. E. , title =. Social Science and Medicine , year =

  79. [79]

    and Liu, L

    Zhang, X. and Liu, L. and Lan, M. and Song, G. and Xiao, L. and Chen, J. , title =. Computers Environment and Urban Systems , year =

  80. [80]

    Geo-spatial Information Science , volume =

    Yuying Wu and Yijing Li , title =. Geo-spatial Information Science , volume =. 2023 , doi =

    2023