{"id":"c4ab32ed-b5c7-49a3-80a5-7b4323cd1eca","arxiv_id":"2506.09731","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Shortest-path stability under small destination shifts varies across cities and urban layouts, with grid-like and central areas being least stable.","lead":"This paper measures how much the shortest driving route changes when the destination is moved by up to 100 meters, across 46 cities worldwide. It finds that central, dense, and grid-like areas produce more variable routes, while stable routes dominate toward the periphery.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Fig. 4 distance trend is collinear with trip length under the radial OD sampling; the exponential fit may be an artifact of path-length scaling.","rationale":"The reader's conditional verdict is appropriate, but the stated weakest assumption (fixed 100 m perturbation across different network granularities) is not the single most load-bearing issue. The headline quantitative result, the universal exponential increase of stability with distance from the city center, is built on an OD sampling scheme in which ring radius and trip length are deterministically coupled, and the stability metric is sensitive to trip length by construction. The paper itself identifies the strong inverse relation between R and stability but does not control for it when presenting Fig. 4. A matched-length or synthetic-network test would settle whether the pattern is a property of urban structure or of the measurement design. I keep the verdict unchanged (still conditional) because the descriptive cross-city comparisons and open-source code remain useful, but the condition should include trip-length stratification before the universal claim is accepted.","tokens_in":15914,"tokens_out":10208,"duration_ms":118135,"concrete_test":"Recompute the Fig. 4 analysis within fixed bins of origin-destination path length l(p(o,d)) (e.g., 1 km bins). For each bin, plot median stability against ring radius r; if the monotone increase disappears or reverses, the distance trend is a trip-length artifact. As a control, run the same pipeline on a homogeneous random geometric graph or square lattice with no distinguished center; if the same exponential r-trend emerges, it cannot diagnose urban periphery. Also report the Spearman correlation between r and median l(p(o,d)) for the sampled OD pairs.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 3.3.1 generates all OD pairs by pairing 36 points on the same concentric circle, so at ring radius r the minimum trip distance is about 0.174r and the median origin-destination chord scales linearly with r. The weighted Jaccard in Section 3.2 is shared/(shared+diverged): for a fixed 100 m destination displacement, the diverged suffix changes little with r, while the shared prefix grows with trip length. Stability therefore rises with r for purely metric reasons, even in a homogeneous network. The paper's own Section 5 Q3 shows R = p(d,d_x)/p(o,d) has Pearson -0.831 with stability; with this sampling design R falls mechanically as r increases. No analysis stratifies by trip length or holds p(o,d) fixed, so the universal 'distance from center' pattern in Fig. 4 and the 'plateau in peripheral areas' interpretation in Section 6 are not separated from OD trip-length scaling. The fixed perturbation-radius issue raised by the reader is real, but it mainly affects cross-city comparisons; the collinearity identified here bears directly on the headline trend.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces a measure of 'path stability' for urban shortest paths under small perturbations of the destination, defined as the average weighted Jaccard similarity between the original path and paths to destinations displaced by Δ=[0,100] m. The method is applied to 46 cities using OSM data, with O-D pairs sampled from points on concentric circles centered on each city. The paper reports a universal exponential increase of stability with distance from the city center, cross-city correlations with average street length, circuity, and orientation entropy, a four-cluster typology of city networks, and spatial autocorrelation of unstable destinations. The work is a substantial extension of the authors' earlier workshop paper and ships open-source code.","tokens_in":16137,"tokens_out":6813,"duration_ms":69864,"significance":"If the central distance-stability trend were robust, the paper would provide a useful empirical descriptor of urban road networks with direct implications for navigation-system design and traffic management. The main strengths are the multi-city dataset, a clearly specified and reproducible measurement pipeline, and the public availability of the code. However, the headline trend is confounded by the O-D sampling design (trip length scales mechanically with ring radius), and cross-city comparisons rely on a fixed perturbation radius that is not commensurate across networks with very different granularity. The descriptive value is real, but the universal claims require additional controls and inferential statistics before they can be accepted.","major_comments":[{"comment":"The radial sampling pairs only points on the same concentric circle, so the expected shortest-path length p(o,d) increases linearly with ring radius r. For a fixed 100 m destination perturbation, the diverged suffix of the path is roughly constant while the shared prefix grows with p(o,d); hence the weighted Jaccard stability increases toward 1 as r grows for purely metric reasons. The exponential fit in Fig. 4 and the Section 6 claim that 'as distance from the city center increases, route stability grows' are therefore not separated from trip-length scaling. Please stratify by trip length or p(o,d) (e.g., regress stability on both radius and trip length, or compare OD pairs of matched length across rings) and report whether the distance-from-center effect persists.","section":"§3.3.1 and Fig. 4"},{"comment":"A fixed perturbation interval Δ=[0,100] m is applied to all cities, but average street length ranges from 67 m (Tokyo) to 213 m (Ottawa). In fine-grained networks a 100 m destination shift can cross many more decision points than in coarse networks, so the cross-city stability rankings and the correlation with average street length (Pearson=0.476, Fig. 7 left) may reflect this granularity mismatch rather than an intrinsic stability property. Please add a robustness check with a normalized perturbation radius (e.g., proportional to city-specific average street length) or report results for several values of Δ.","section":"§4 and Table 4"},{"comment":"The city-level results are reported as Pearson correlations and an exponential R² without confidence intervals, p-values, or multiple-comparison correction. For n=46 cities, correlations around 0.4 are only marginally significant, and the word 'significant' in Q4/Q5 is not supported by inferential statistics. Please provide bootstrap confidence intervals or p-values for all reported correlations and fit parameters, and consider a mixed-effects model that accounts for city and ring effects.","section":"§5 (Figs. 4, 6, 7)"}],"minor_comments":[{"comment":"The phrase 'is this study' should be 'in this study'.","section":"Introduction"},{"comment":"'realtive' should be 'relative'.","section":"Section 5 (introductory paragraph)"},{"comment":"'Moskow' should be 'Moscow'.","section":"Table 4"},{"comment":"The text gives fit parameters a=−0.26, b=−0.43, c=0.99, but the caption's equation is garbled (it appears to omit the negative exponent and the sign of a); please correct the equation for clarity.","section":"Fig. 4"},{"comment":"The city-shape classification (Grid/Radial/Organic) is based on visual inspection; please include a reproducibility note or an inter-rater consistency check.","section":"Section 5, Q5"},{"comment":"The entropy-of-bearing computation is cited to [27] but the binning and calculation details are not described; a short description would improve reproducibility.","section":"Section 5, Q5"}],"recommendation":"major_revision","confidential_remarks":"The title and framing emphasize a universal law, but the supporting evidence is currently confounded by the sampling scheme. The authors should be pushed to perform the trip-length stratification; if the effect disappears, the manuscript's contribution reduces to a descriptive cross-city comparison, which may still be publishable with appropriately weakened claims. I have no concerns about citation practices beyond a modest amount of self-citation, and the prior workshop paper is properly acknowledged."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First thing you should know: this is a careful descriptive study with open-source code, but its central claim—stability grows with distance from the city center—is probably an artifact of how the OD pairs are generated. The authors sample origins and destinations on the same concentric circle, so trip length scales with ring radius. For a fixed 100 m destination displacement, the diverged portion of the route stays roughly constant while the shared prefix grows with trip length, making the weighted Jaccard increase with radius. Their own Q3 result, R = p(d,d_x)/p(o,d) with Pearson -0.831 against stability, is exactly this mechanical relationship. They mention short OD distances in central areas in the Discussion, but they never control for trip length. The exponential fit in Fig. 4 (R²=0.98) describes the chord-length distribution more than any property of city structure.\n\nWhat is genuinely useful: the stability measure is clearly defined, the pipeline is reproducible, and the 46-city comparison is a real extension of the Barcelona workshop paper. The city-level correlations with street length and circuity are plausible, and the discussion of stability as neutral—flexibility vs. structural constraint—is thoughtful.\n\nSoft spots beyond the confound: the perturbation radius Δ=[0,100] m is fixed across cities whose street lengths differ by about 3x; in Tokyo 100 m crosses several intersections, in Ottawa it may be one block. Any cross-city comparison of stability is entangled with this scale mismatch. The correlations are also reported without confidence intervals or p-values; at n=46, a Pearson of 0.25 (entropy) is not significant. The shape classification is subjective, though they admit it.\n\nMy take: the paper deserves a serious referee, because the idea is good and the empirical effort is substantial. But the main result needs reanalysis. I would ask for: (1) a trip-length-stratified version of Fig. 4, holding p(o,d) fixed within bins or using OD pairs across rings; (2) normalization of Δ by local block or street length; (3) significance tests for the correlations. If the distance trend survives trip-length control, this becomes a strong comparative paper. If not, the cross-city results may still stand, but the headline must be rewritten.\n\nI would send it to review with a request for these analyses. Worth reading for the methodology, but do not cite the distance trend yet.","headline":"Useful empirical extension with reproducible code, but the headline distance-stability trend is likely a trip-length scaling artifact of the radial OD sampling—deserves peer review with mandatory reanalysis.","tokens_in":16648,"tokens_out":3906,"would_cite":false,"duration_ms":43256,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Small destination shifts can rewrite a route in any city, most strongly near the center.","keywords":["shortest path","path stability","road network","urban navigation","destination perturbation","route variability","city comparison","urban structure"],"falsifier":"Recompute the analysis with perturbation radii of 50 m, 200 m, and 500 m, and also with the radius scaled to each city's average street length; if the cross-city stability ranking or the exponential center-periphery curve changes substantially under rescaling, the fixed 100-meter band is the real driver of the result. Alongside that, a single city or district where median stability does not increase from center to periphery would break the claimed universal trend.","tokens_in":15704,"feed_emoji":"🧭","tokens_out":7523,"duration_ms":79731,"temperature":0.7,"pith_summary":"This paper asks whether the common experience of navigation apps proposing completely different routes for two nearby destinations is a quirk of one city or a general property of urban road networks. It defines path stability—how much the shortest route to a destination changes when the destination is moved by at most 100 meters—and measures it for 46 cities using destination perturbations and a length-weighted Jaccard similarity between original and rerouted paths. The paper's central claim is that the pattern is universal but heterogeneous: stability increases with distance from the city center and plateaus near total stability in peripheral areas, following $y = 0.26e^{-0.43x} + 0.99$ with $R^2 = 0.98$. At the same time, cities divide into stable and unstable profiles, with unstable cities tending to have grid-like, fine-grained networks and longer detours between a destination and its perturbed neighbors. If the claim holds, route sensitivity to small destination shifts is not noise but a measurable, city-specific property tied to network geometry, with concrete uses in traffic distribution and navigation design.","feed_headline":"Route stability grows with distance from city center in all 46 cities","feed_subtitle":"A 100-meter shift in destination rewires downtown routes but barely changes trips beyond 10 kilometers.","key_machinery":"The central object is the path-stability score $S^{\\Delta}(o,d)$ for an origin-destination pair: the destination is displaced by at most $\\Delta = [0,100]$ meters, candidate displaced nodes are sampled in eight sectors of a ring around it, shortest paths are recomputed, and the average length-weighted Jaccard index between the original and displaced routes gives the score. The same score aggregated by radius bands and by city carries the analysis: the exponential fit $y = 0.26e^{-0.43x} + 0.99$ formalizes the center-periphery trend, while the normalized detour ratio $R$ and street-level indicators such as average street length, circuity, and bearing entropy explain cross-city differences.","core_discovery":"The core discovery, stated on the paper's own terms, is that shortest-path instability under small destination perturbations is a universal but heterogeneous property of cities. Across all 46 cities studied, median stability increases with radial distance from the city center and plateaus near 1 beyond roughly 10 km; the aggregate curve is $y = 0.26e^{-0.43x} + 0.99$ with $R^2 = 0.98$. Cross-city differences align with network structure: cities with grid-like layouts, short average street segments, low orientation entropy, and large normalized detours between a destination and its displaced neighbors rank as unstable, while stable cities have longer, more varied segments and radial or organic layouts. The normalized detour ratio, comparing the path length from the original destination to a displaced one against the full trip length, is the strongest single correlate of instability at city level, with Pearson correlation $-0.831$.","pith_inferences":["Because a fixed 100-meter shift crosses more decision points in fine-grained grids, the stability ranking across cities may partly encode network granularity rather than intrinsic route robustness; scaling the perturbation by local street length would test this directly.","The mechanism suggests a local, testable prediction: adding short cross-streets or increasing intersection density near a destination should lower its stability score, while removing connections should raise it, so the score could serve as a response variable in network-design experiments.","The paper treats stability as a descriptive lens, but the same score could be operationalized in routing: a navigation service could deliberately route through unstable zones during events to spread traffic, or keep to stable corridors when concentration on main arteries is desired.","Within-city maps show unstable destinations cluster together; a formal spatial-statistical test of that autocorrelation, beyond visual inspection, would quantify how localized the phenomenon is."],"forward_implications":["Navigation providers can produce per-destination stability maps, flagging places where a user's small destination adjustment will likely produce a completely different route.","Traffic managers can use low-stability zones around event venues to disperse incoming vehicles across multiple routes instead of funneling them down one path.","The center-periphery curve gives a default expectation: beyond roughly 10 km from the city center, small destination shifts should barely change the recommended shortest path.","Road-network metrics—average street length, variation in segment length, circuity, and orientation entropy—can be used to predict whether a city or district will be route-stable before any routing simulation is run."],"supporting_citations":[{"why":"Supplies the prior Barcelona observation and the first formulation of path instability that this paper revises and extends.","marker":"[2]"},{"why":"Provides the fixed-radius radial sampling approach used to generate origin-destination pairs at increasing distances from the city center.","marker":"[4]"},{"why":"Anchors the radial sampling methodology in established studies of travel-route morphology and urban organization.","marker":"[10]"},{"why":"Defines the weighted Jaccard index used to quantify route overlap between original and perturbed paths.","marker":"[26]"},{"why":"Provides the bearing-entropy method used to measure the grid-likeness of street orientation.","marker":"[27]"},{"why":"Supplies the Grid/Radial/Organic classification of urban form used in the shape analysis.","marker":"[28]"}],"fun_headline_variants":["Destination tweaks reroute downtown, not suburbs, in 46 cities","Grid cities have unstable routes under small destination changes","Route stability: distance from center predicts fragility","Study: city layout determines route robustness to destination shifts","Path stability under perturbation varies by city structure"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that a 100-meter displacement of a destination counts as the same 'small change' in every city, even though average street length runs from 67 meters in Tokyo to 213 meters in Ottawa, so a fixed shift crosses very different numbers of decision points in different networks.","fun_headline_variants_meta":{"raw":{"variants":["Destination tweaks reroute downtown, not suburbs, in 46 cities","Grid cities have unstable routes under small destination changes","Route stability: distance from center predicts fragility","Study: city layout determines route robustness to destination shifts","Path stability under perturbation varies by city structure"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000334,"raw_usage":{"total_tokens":1820,"prompt_tokens":876,"completion_tokens":944,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":492,"completion_tokens_details":{"reasoning_tokens":869}},"tokens_in":492,"tokens_out":944,"duration_ms":11718,"temperature":1.0,"reasoning_tokens":869,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T04:42:29.529506+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Recompute the analysis with perturbation radii of 50 m, 200 m, and 500 m, and also with the radius scaled to each city's average street length; if the cross-city stability ranking or the exponential center-periphery curve changes substantially under rescaling, the fixed 100-meter band is the real driver of the result. Alongside that, a single city or district where median stability does not increase from center to periphery would break the claimed universal trend.","supporting_citations":[{"cited_title":"In: EDBT/ICDT Workshops","cited_arxiv_id":null,"evidence_quote":"Supplies the prior Barcelona observation and the first formulation of path instability that this paper revises and extends."},{"cited_title":"Nature communications8(1), 2229 (2017)","cited_arxiv_id":null,"evidence_quote":"Anchors the radial sampling methodology in established studies of travel-route morphology and urban organization."},{"cited_title":"Applied Network Science4(1), 1–19 (2019)","cited_arxiv_id":null,"evidence_quote":"Provides the bearing-entropy method used to measure the grid-likeness of street orientation."},{"cited_title":"[Online; accessed 2024-10-29] (2024)","cited_arxiv_id":null,"evidence_quote":"Supplies the Grid/Radial/Organic classification of urban form used in the shape analysis."}],"review_version":1}