arxiv: 2605.10522 · v1 · submitted 2026-05-11 · 💻 cs.HC

Recognition: no theorem link

The Balance between Nuance and Clarity: Decluttering Tabular Sequential Graphs to Counter Money Laundering

Louise Fallon, Pedro Bizarro, Rita Costa, Salom\'e Esteves

Authors on Pith no claims yet

Pith reviewed 2026-05-12 04:58 UTC · model grok-4.3

classification 💻 cs.HC

keywords money launderinggraph visualizationsequential graphsnode reductionuser studytransaction groupingfinancial analysisdecluttering

0 comments

The pith

Grouping transactions by amount and timing reduces clutter in sequential graphs used for money laundering analysis, with a user study showing that simpler reductions are not always the most useful.

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

The paper presents a tabular sequential graph visualization designed to track money flows starting from an alerted victim account across accounts and banks. It introduces three grouping methods to shrink the number of nodes and edges: one based on transaction amounts, one based on timing, and one that combines both amount and sequence order. A study with experts found that the grouping which cut the most elements was not always the one experts preferred for spotting patterns. This work shows the practical tension between making graphs easier to scan quickly and keeping enough detail for thorough investigation of suspicious flows.

Core claim

The authors propose tabular sequential graphs that follow transaction sequences row by row by bank and introduce three explicit grouping methods—amount-based, time-based, and combined amount-and-order—to declutter them. Their expert user study demonstrates that greater node reduction does not automatically improve analytical value and that more granular versions trade off extra manual effort against longer interpretation time.

What carries the argument

Tabular sequential graph with rows as banks and the three grouping methods (amount-based, time-based, combined amount-and-order) that merge nodes and edges while preserving transaction sequences.

If this is right

Analysts gain a clearer overview of money flows with fewer elements to examine manually.
The combined grouping method provides a middle path that reduces clutter without fully sacrificing sequence information.
More granular graphs increase the time needed for interpretation even though they reduce the need for additional manual grouping steps.
Daily alert investigation benefits from choosing groupings that match the required level of pattern visibility rather than maximum simplification.

Where Pith is reading between the lines

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

The same decluttering approach could be adapted to visualize other sequential transaction problems such as fraud rings or sanctions evasion.
Tools built on these methods might benefit from letting analysts switch groupings on the fly based on the specific alert under review.
The observed trade-off implies that visualization systems for investigative work should measure both reduction metrics and task completion time rather than reduction alone.
Larger studies with analysts from different institutions could test whether the preference patterns hold beyond the original participants.

Load-bearing premise

The grouping methods keep enough transaction details intact for accurate money laundering detection and the expert participants' preferences represent typical operational needs.

What would settle it

A controlled test where analysts using each grouped version fail to detect known laundering chains that remain visible only in the ungrouped graph.

Figures

Figures reproduced from arXiv: 2605.10522 by Louise Fallon, Pedro Bizarro, Rita Costa, Salom\'e Esteves.

**Figure 1.** Figure 1: Example of a tabular sequential graph visualization for a synthetic money laundering network before and after the [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗

**Figure 2.** Figure 2: The tabular sequential network of synthetic data of a case [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗

**Figure 4.** Figure 4: A meta-node encoding a group of four accounts. [PITH_FULL_IMAGE:figures/full_fig_p003_4.png] view at source ↗

**Figure 5.** Figure 5: Combined node grouping in a synthetic network similar to [PITH_FULL_IMAGE:figures/full_fig_p003_5.png] view at source ↗

**Figure 6.** Figure 6: Time-based node grouping in a synthetic network similar to [PITH_FULL_IMAGE:figures/full_fig_p004_6.png] view at source ↗

read the original abstract

Money laundering is not only about moving illicit funds, but about hiding the money's origin and traces to complicate detection. Financial criminals resort to many methods to avoid regulators and legal thresholds. But analysts investigating alerts, dedicated to pin mule accounts and track suspicious transactions daily, also have theirs. Network visualizations can be key in countering adversarial money laundering activities, especially if they provide a clear overview of the money flows and a seamless analysis experience, but they are often not structured for this type of task. That is why we propose a tabular sequential graph visualization tailored to money laundering analysis - following transactions (edges) from the victim account that triggered an alert through multiple accounts (nodes) and banks (rows). To reduce the number of nodes and edges, we propose three methods for grouping these tabular sequential graphs: an amount-based approach, a time-based approach, and a combined solution that considers both the transaction amount and its order. A user study with experts revealed that the most effective method in node reduction was not necessarily the most interesting for analysis and that there is a trade-off between manual work and time for interpretation in more granular graphs.

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.

Desk Editor's Note private letter to a colleague

The paper gives a practical set of grouping rules for tabular transaction graphs in AML work, but the user study is described too thinly to back the claimed trade-off between reduction and usefulness.

read the letter

The main takeaway is that this work adapts graph decluttering to the specific needs of money laundering analysts by laying out transactions in a tabular sequential format and then collapsing nodes with three simple rules: group by similar amounts, by close timestamps, or by both. That format follows the flow from an alert account through banks and mule accounts, which matches how analysts actually trace patterns rather than generic network drawings. The grouping ideas are straightforward and directly tied to the domain, so they could be implemented without much overhead. Credit for spotting that standard visualizations overload analysts with nodes while still needing to keep order and value details intact. The paper does a clean job naming the tension between overview and detail in this setting. The soft spot is the evaluation. The abstract reports that experts found the strongest reducer not always the most interesting and noted a manual-work versus interpretation-time trade-off, yet it supplies no participant numbers, task descriptions, metrics, or statistical checks. Without those, it is hard to tell whether the groupings actually preserve the timing or amount anomalies that matter for detection or whether the opinions would hold for other teams. The stress-test note is right on this point. If the full paper adds the protocol and raw results, the claim becomes testable; as presented, the evidence is light. This is aimed at HCI people who build tools for compliance teams or at AML practitioners looking for visualization tweaks. A reader in either group could borrow the grouping logic even if the study needs bolstering. It deserves peer review because the problem is real, the methods are concrete enough to replicate, and referees can push for the missing study details that would make the trade-off claim solid.

Referee Report

2 major / 1 minor

Summary. The paper proposes tabular sequential graphs to visualize money laundering transaction flows from an alert account through accounts and banks. It introduces three node-grouping methods (amount-based, time-based, and combined amount-and-order) to reduce clutter, and reports a user study with experts finding that the most effective reduction method is not necessarily the most analytically interesting, with a noted trade-off between manual effort and interpretation time in granular graphs.

Significance. If the groupings can be shown to retain laundering-relevant details and the user study is properly documented with objective metrics, the work could inform visualization design for financial analysts by addressing the nuance-clarity tension in sequential data. The focus on expert feedback in a real domain is a positive aspect.

major comments (2)

[Abstract] Abstract and user study description: The central empirical claim (that the user study revealed a trade-off between node reduction effectiveness and analytical interest) is unsupported because no details are supplied on the number of experts, their qualifications, the specific tasks or graphs used, the metrics for 'effectiveness' and 'interest' (e.g., time, accuracy, Likert scores), or any statistical tests. This leaves the reported findings unverifiable and the main contribution unsubstantiated.
[Grouping Methods] Grouping methods section: The three methods are introduced at a conceptual level without explicit algorithms, decision thresholds (e.g., amount similarity or time window criteria), or examples demonstrating preservation of sequential order, amounts, and bank rows. It is therefore impossible to verify whether critical anomalies (such as timing or value patterns used to trace mule accounts) are retained, which is load-bearing for the claim that the methods support accurate detection while reducing nodes.

minor comments (1)

The abstract and introduction introduce specialized terms such as 'tabular sequential graph' and the three grouping methods without a concise definition or motivating example at first use; adding a short illustrative figure or paragraph would improve accessibility.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback, which highlights important areas for improving the verifiability of our contributions. We address each major comment below and will revise the manuscript to provide greater detail and transparency.

read point-by-point responses

Referee: [Abstract] Abstract and user study description: The central empirical claim (that the user study revealed a trade-off between node reduction effectiveness and analytical interest) is unsupported because no details are supplied on the number of experts, their qualifications, the specific tasks or graphs used, the metrics for 'effectiveness' and 'interest' (e.g., time, accuracy, Likert scores), or any statistical tests. This leaves the reported findings unverifiable and the main contribution unsubstantiated.

Authors: We agree that the abstract is overly concise and does not provide enough information to substantiate the user-study claims. The full manuscript includes a dedicated user-study section with descriptions of the expert participants, tasks, and qualitative observations on the trade-off. To address the concern, we will expand the abstract to include key details (number of experts, their qualifications as financial analysts, tasks performed, metrics such as Likert-scale ratings for analytical interest and perceived effectiveness, and any statistical tests applied) and ensure the user-study section supplies complete, verifiable documentation of all elements. revision: yes
Referee: [Grouping Methods] Grouping methods section: The three methods are introduced at a conceptual level without explicit algorithms, decision thresholds (e.g., amount similarity or time window criteria), or examples demonstrating preservation of sequential order, amounts, and bank rows. It is therefore impossible to verify whether critical anomalies (such as timing or value patterns used to trace mule accounts) are retained, which is load-bearing for the claim that the methods support accurate detection while reducing nodes.

Authors: We acknowledge that the grouping-methods section remains at a high conceptual level. In the revision we will add explicit algorithmic descriptions, concrete decision thresholds (e.g., amount similarity within a defined percentage and time windows of a specified duration), pseudocode or step-by-step procedures, and worked examples that illustrate retention of sequential order, transaction amounts, and bank rows. These additions will allow readers to confirm that laundering-relevant patterns, such as timing and value anomalies for tracing mule accounts, are preserved. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical user study with independent evaluation

full rationale

The paper proposes three grouping methods (amount-based, time-based, combined) for decluttering tabular sequential graphs and evaluates their effectiveness via a user study with domain experts. No mathematical derivations, equations, fitted parameters, or predictions are present that could reduce to inputs by construction. Claims about node reduction effectiveness and trade-offs between granularity and interpretability rest directly on the external user study results rather than any self-referential loop or self-citation chain. This is standard empirical HCI work with no load-bearing internal circularity.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 4 invented entities

The paper introduces a new visualization format and three decluttering methods whose value is assessed via expert feedback; no mathematical free parameters or derivations are present.

axioms (1)

domain assumption Expert user feedback in a controlled study provides reliable signals about visualization utility for domain tasks
The central claim about trade-offs depends on the validity of the user study results.

invented entities (4)

Tabular sequential graph no independent evidence
purpose: Linear row-per-bank layout for following transaction sequences in AML alerts
Core new visualization structure proposed to address clutter in standard network graphs.
Amount-based grouping method no independent evidence
purpose: Declutter by aggregating transactions according to amount similarity
One of the three proposed reduction techniques.
Time-based grouping method no independent evidence
purpose: Declutter by aggregating transactions according to temporal proximity
One of the three proposed reduction techniques.
Combined amount-and-order grouping method no independent evidence
purpose: Declutter using both transaction amount and sequence order
One of the three proposed reduction techniques.

pith-pipeline@v0.9.0 · 5505 in / 1315 out tokens · 56982 ms · 2026-05-12T04:58:49.108356+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

25 extracted references · 25 canonical work pages

[1]

Cambridge Intelligence: data visualization technology. 2

work page
[2]

DataWalk: Graph and AI-enabled Analysis, Intelligence and Investi- gation Platform. 2

work page
[3]

Neo4j Graph Intelligence Platform, Apr. 2026. 2

work page 2026
[4]

B. Bach, N. H. Riche, C. Hurter, K. Marriott, and T. Dwyer. Towards Unambiguous Edge Bundling: Investigating Confluent Drawings for Network Visualization.IEEE transactions on visualization and com- puter graphics, 23(1):541–550, Jan. 2017. doi: 10.1109/TVCG.2016. 2598958 2

work page doi:10.1109/tvcg.2016 2017
[5]

Balzer and O

M. Balzer and O. Deussen. Level-of-detail visualization of clustered graph layouts. In2007 6th International Asia-Pacific Symposium on Visualization, pp. 133–140. IEEE, Sydney, NSW, Feb. 2007. doi: 10. 1109/APVIS.2007.329288 2

work page arXiv 2007
[6]

U. Brandes. Drawing on Physical Analogies. In G. Goos, J. Hart- manis, J. Van Leeuwen, M. Kaufmann, and D. Wagner, eds.,Draw- ing Graphs, vol. 2025, pp. 71–86. Springer Berlin Heidelberg, Berlin, Heidelberg, 2001. Series Title: Lecture Notes in Computer Science. doi: 10.1007/3-540-44969-8 4 1

work page doi:10.1007/3-540-44969-8 2025
[7]

Contreras and E

A. Contreras and E. Villa P ´erez. Strategic choices in money launder- ing: Smurfing, layering, and financial over-diversification.Journal of Economic Criminology, 11:100199, Mar. 2026. doi: 10.1016/j.jeconc .2025.100199 1

work page doi:10.1016/j.jeconc 2026
[8]

W. Dong, X. Fu, G. Xu, and Y . Huang. An improved force-directed graph layout algorithm based on aesthetic criteria.Computing and Visualization in Science, 16(3):139–149, June 2013. doi: 10.1007/ s00791-014-0228-5 1

work page 2013
[9]

D. Edge, J. Larson, M. Mobius, and C. White. Trimming the Hairball: Edge Cutting Strategies for Making Dense Graphs Usable. In2018 IEEE International Conference on Big Data (Big Data), pp. 3951–

work page
[10]

IEEE, Seattle, W A, USA, Dec. 2018. doi: 10.1109/BigData. 2018.8622521 2

work page doi:10.1109/bigdata 2018
[11]

B. E. Feliciano.Increasing Readability of Large Networks Visualiza- tions. PhD thesis, Instituto Superior T´ecnico, Universidade de Lisboa,

work page
[12]

Höllt, N

V . Filipov, A. Arleo, and S. Miksch. Are We There Yet? A Roadmap of Network Visualization from Surveys to Task Taxonomies.Com- puter Graphics Forum, 42(6):e14794, Sept. 2023. doi: 10.1111/cgf. 14794 2

work page doi:10.1111/cgf 2023
[13]

P. M. Gilmour. Reexamining the anti-money-laundering framework: a legal critique and new approach to combating money laundering. Journal of Financial Crime, 30(1):35–47, Mar. 2022. doi: 10.1108/ JFC-02-2022-0041 1

work page 2022
[14]

Halford, I

E. Halford, I. Gibson, M. Newfield, and M. Dhanwala. Developing a scoring model for managing money laundering transactions using machine learning.Journal of Money Laundering Control, 28(7):30– 49, Mar. 2025. doi: 10.1108/JMLC-09-2024-0152 1

work page doi:10.1108/jmlc-09-2024-0152 2025
[15]

Huang, P

W. Huang, P. Eades, and S.-H. Hong. Measuring Effectiveness of Graph Visualizations: A Cognitive Load Perspective.Information Vi- sualization, 8(3):139–152, Sept. 2009. Publisher: SAGE Publications. doi: 10.1057/ivs.2009.10 1

work page doi:10.1057/ivs.2009.10 2009
[16]

Huang, P

W. Huang, P. Eades, S.-H. Hong, and C.-C. Lin. Improving multiple aesthetics produces better graph drawings.Journal of Visual Lan- guages & Computing, 24(4):262–272, Aug. 2013. doi: 10.1016/j.jvlc .2011.12.002 1

work page doi:10.1016/j.jvlc 2013
[17]

H. C. Purchase, D. Carrington, and J.-A. Allder. Empirical Evaluation of Aesthetics-based Graph Layout.Empirical Software Engineering, 7(3):233–255, Sept. 2002. doi: 10.1023/A:1016344215610 2

work page doi:10.1023/a:1016344215610 2002
[18]

Saket, P

B. Saket, P. Simonetto, S. Kobourov, and K. B ¨orner. Node, Node- Link, and Node-Link-Group Diagrams: An Evaluation.IEEE Trans- actions on Visualization and Computer Graphics, 20(12):2231–2240, Dec. 2014. doi: 10.1109/TVCG.2014.2346422 2

work page doi:10.1109/tvcg.2014.2346422 2014
[19]

Shadrooh and K

S. Shadrooh and K. Nørv ˚ag. SMoTeF: Smurf money laundering de- tection using temporal order and flow analysis.Applied Intelligence, 54(15):7461–7478, Aug. 2024. doi: 10.1007/s10489-024-05545-4 1

work page doi:10.1007/s10489-024-05545-4 2024
[20]

Shneiderman and C

B. Shneiderman and C. Dunne. Interactive Network Exploration to Derive Insights: Filtering, Clustering, Grouping, and Simplification. In W. Didimo and M. Patrignani, eds.,Graph Drawing, pp. 2–18. Springer, Berlin, Heidelberg, 2013. doi: 10.1007/978-3-642-36763 -2 2 2

work page doi:10.1007/978-3-642-36763 2013
[21]

Teichmann

F. Teichmann. Recent trends in money laundering.Crime, Law and Social Change, 73(2):237–247, Mar. 2020. doi: 10.1007/s10611-019 -09859-0 1

work page doi:10.1007/s10611-019 2020
[22]

Vehlow, F

C. Vehlow, F. Beck, and D. Weiskopf. Visualizing Group Structures in Graphs: A Survey.Computer Graphics Forum, 36(6):201–225, 2017. eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1111/cgf.12872. doi: 10.1111/cgf.12872 2

work page doi:10.1111/cgf.12872 2017
[23]

Y . Wang, G. Baciu, and C. Li. Visualizing the Temporal Similarity Between Clusters of Dynamic Graphs. In2019 IEEE 18th Interna- tional Conference on Cognitive Informatics & Cognitive Computing (ICCI*CC), pp. 205–210, July 2019. doi: 10.1109/ICCICC46617. 2019.9146098 2

work page doi:10.1109/iccicc46617 2019
[24]

Ware and R

C. Ware and R. Bobrow. Motion to support rapid interactive queries on node–link diagrams.ACM Trans. Appl. Percept., 1(1):3–18, July

work page
[25]

doi: 10.1145/1008722.1008724 1

work page doi:10.1145/1008722.1008724