{"id":"2dd43aba-e9fc-4359-abd5-7a5d4c90f5a6","arxiv_id":"1908.05929","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Ride-sharing efficiency on any street network collapses onto the universal curve E = Emax B/(B+B1/2), with B1/2 encoding how hard the network's shortest-path structure makes ride-sharing.","lead":"This paper finds that the efficiency of ride-sharing fleets, measured by scheduled customers per vehicle, follows one universal curve across many street network types after rescaling by a single 'topological factor'. If the finding holds, planners could estimate how many vehicles a new city or rural area needs to reach a target service level.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The B^{-1} waiting-time scaling in Eq. (5) is assumed, not derived, and the universality claim rests entirely on it; the paper's own conclusion concedes this scaling need not hold for other dispatchers.","rationale":"The reader's weakest_assumption is exactly the B^{-1} waiting-time scaling of Eq. (5), and the paper's own conclusion flags it. My independent reading confirms that this is the load-bearing step: it is the only place where the functional form f(z) = 1/(1+z^{-1}) enters, and it is neither derived nor empirically isolated. The supplementary material partially addresses the concern by showing fits to the asymptotic form, but the fits are per-network and the exponent is not reported—only the inferred B1/2 values. The paper also contains a self-acknowledged limitation: the conclusion states that the universality is not guaranteed for dispatchers with different scaling, which directly contradicts the abstract's 'insensitive to ... dispatching criteria' claim. The supplement (Fig. S4) resolves the contradiction only by redefining universality as 'same functional form with different parameters per dispatcher,' which weakens the central claim. That said, the empirical collapse for the implemented dispatchers is a real result; the two-parameter collapse is nontrivial for 14 networks. The concern is therefore not that the data are wrong, but that the universal curve's shape rests on an unverified scaling law. The proposed concrete test—directly measuring the <tw> exponent and testing a dispatcher with different scaling—would settle whether the universality is a genuine law or an artifact of the dispatcher class. I agree with the reader's conditional verdict: the paper is publishable with this caveat, but the universality claim should be qualified until the exponent is directly verified and the dispatcher-dependence is squarely addressed.","tokens_in":17042,"tokens_out":1020,"duration_ms":10839,"concrete_test":"Re-analyze the raw simulation data to extract <tw>(B) at fixed high x for each network (the data behind Fig. S1(b)) and fit log <tw> versus log B over the asymptotic B range. If the fitted exponent is statistically consistent with -1 with a constant offset (i.e., <tw> = gamma*tau*B^{-1}) for all networks, the derivation of f(z) holds. Additionally, run one controlled simulation with a deliberately 'non-B^{-1}' dispatcher, e.g., one that inserts new requests only at the end of a bus's current route (so the wait time is set by the bus's remaining route, not by headway), and check whether the efficiency curve still collapses to 1/(1+z^{-1}). If the exponent deviates from -1 or the collapse fails, the universal curve is dispatcher-specific, not universal.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central claim E = Emax f(B/B1/2) with universal f(z) = 1/(1+z^{-1}) is derived in Eqs. (4)-(7) from the assumed asymptotic scalings <td> ~ <l>/v ~ B^0 and <tw> ~ gamma*tau*B^{-1}. The drive-time scaling is well supported by the simulation data in Fig. S1(a), but the waiting-time scaling is introduced heuristically ('twice as many buses means a bus going in the right direction comes by twice as often') rather than derived from the dispatcher dynamics. The proportionality constant gamma is then fitted per network and identified with B1/2, so the functional form f(z) = 1/(1+z^{-1}) is not independently tested: the collapse in Fig. 3(a,b) is a two-parameter fit per network to a curve whose shape is fixed by the assumed B^{-1} law. The paper's own concluding paragraph concedes that 'the same asymptotic universality is not guaranteed to hold for hypothetical dispatchers with a different scaling.' Since the main text claims the law is 'insensitive to ... dispatching criteria' while the supplement (Fig. S4) shows that different dispatchers give different universal functions f(·), the claimed universality across dispatchers is overstated: what is universal is only that the data can be fit to the same functional form with per-dispatcher parameters, not that the shape is dispatcher-independent. A stronger test would directly measure <tw>(B) in simulations and check the exponent, and would test a dispatcher whose waiting time provably scales differently.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces an efficiency measure E = lim_{x→∞} (⟨C⟩/x)^{-1} for ride-sharing fleets, where ⟨C⟩ is the mean number of scheduled customers per bus and x the normalized request rate. Using a Little's-law-type identity in Eq. (3), the authors connect ⟨C⟩ to mean waiting and driving times, and from the asymptotic assumptions ⟨td⟩ ∝ B^0 and ⟨tw⟩ ∝ B^{-1} they derive the scaling law E = Emax f(B/B1/2) with f(z)=1/(1+z^{-1}). They report that this law collapses simulation data across model topologies and empirical street networks, that B1/2 is controlled by the distinctness of shortest paths, and that the law is insensitive to request distributions and dispatching criteria. The paper concludes by acknowledging that the universality is conditional on the assumed waiting-time scaling and on the absence of fleet-generated congestion.","tokens_in":17371,"tokens_out":9177,"duration_ms":79640,"significance":"If the central scaling law Eq. (2) holds as stated, the paper would supply a practical two-parameter characterization of ride-sharing efficiency across cities and rural areas, with a clean Little's-law foundation in Eq. (3) and a plausible topological interpretation for B1/2. The authors deserve credit for testing a wide range of model and empirical street networks, for explicitly measuring the asymptotic drive-time and waiting-time scalings in Fig. S1, and for honestly acknowledging the limits of the universality claim in the conclusion. However, the strength of the reported universality is currently limited by the fact that the functional form is derived from an assumed B^{-1} waiting-time scaling, the collapse is obtained after per-network fitting of B1/2 and Emax, and the cross-dispatcher universality is contradicted by the paper's own supplementary material. These issues are correctable and do not invalidate the core identity, but they do require a revised and more guarded statement of the claim.","major_comments":[{"comment":"The universal functional form f(z)=1/(1+z^{-1}) in Eq. (7) is derived entirely from the assumed asymptotic scaling ⟨tw⟩ ∼ γτ B^{-1} in Eq. (5) together with the constant drive-time scaling in Eq. (4). The drive-time scaling is well supported by Fig. S1(a), but the waiting-time scaling is introduced heuristically and not derived from the dispatcher dynamics. Because the proportionality constant γ is then identified with the fitted half-efficiency fleet size B1/2, the collapse in Fig. 3(a,b) is a two-parameter fit per network to a curve whose shape is fixed by the assumed B^{-1} law; it does not independently test the universal shape. I recommend directly fitting the exponent of ⟨tw⟩(B) from the simulation data, reporting residuals of the collapse, and demonstrating that a different plausible scaling (e.g., B^{-1/2}) leads to a different f before claiming parameter-free universality.","section":"Scaling of ride-sharing efficiency"},{"comment":"The abstract and the section 'Topological universality' state that the scaling is 'insensitive to ... dispatching criteria,' but Supplement Fig. S4 states that different dispatcher algorithms follow different universal functions f(·), and the conclusion concedes that the same asymptotic universality 'is not guaranteed to hold for hypothetical dispatchers with a different scaling.' These statements are in tension. The universality claim should be restricted to the class of dispatchers for which the asymptotic drive time is constant and the waiting time scales as B^{-1}; otherwise the central claim is overstated.","section":"Topological universality"},{"comment":"Figure 5 presents the relation B1/2 vs. ltot/⟨l⟩ as showing a 'strong dependence,' but the line is only a guide to the eye; no regression statistics, error bars on the fit, or residuals are given. The ratio ltot/⟨l⟩ is an ad-hoc measure that may conflate network size and topology, so without a quantitative model this secondary claim is not established. If this relation is intended as a predictive formula for B1/2, it needs a regression with confidence bounds and out-of-sample validation.","section":"Distinctness of shortest paths controls scaling factor"},{"comment":"The empirical collapse in Fig. 3(b) is computed at x=2.5 and the model collapse in Fig. 3(a) at x=7.5, while E is defined in Eq. (1) as the x→∞ limit of (⟨C⟩/x)^{-1}. The universal curve in Eq. (7) is an asymptotic result for large B and large x; using E measured at a finite, and network-dependent, load makes the collapse a statement about finite-load efficiency, not about the asymptotic law. Please demonstrate that E(B) is insensitive to x in the range used, or fit E at a common, sufficiently large x for all networks, and report how the fitted B1/2 and Emax depend on x.","section":"Efficiency of ride-sharing"}],"minor_comments":[{"comment":"Equation (S1) appears to contain an extra factor of 2 in the expression for ⟨C⟩ compared with Eq. (3) of the main text; please verify that this is a typographical error.","section":"Supplementary Material"},{"comment":"The caption of Fig. S4 says that different dispatchers 'follow different universal functions f(·)' while the main text claims a single universal f; this wording is confusing and should be reconciled.","section":"Supplementary Material"},{"comment":"Reference [1] contains a spelling error ('Sustainbale' instead of 'Sustainable').","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The manuscript would be strengthened if the authors provided quantitative collapse metrics (e.g., residuals) and made the simulation code public for reproducibility; the editor may wish to request these during revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThis paper is worth a serious look. It introduces a clean efficiency measure (average scheduled customers per bus at fixed normalized load), shows a data collapse E = Emax f(B/B1/2) with f(z) = z/(1+z) across a wide range of model and empirical street networks, and backs it with a simple mean-field calculation. Eq. (3) is just Little's law, and the B1/2 values for minimal, ring, complete, and star graphs match simulations without free fitting beyond the model. The empirical collapse across islands, cities, and rural areas is genuinely useful for planning fleet sizes in new regions.\n\nThe soft spots are real but not fatal. The derivation of f depends on the assumed B^{-1} waiting-time scaling, Eq. (5). The stress-test note says this is assumed, but check the supplement: Fig. S1(b) directly plots <tw> vs B and confirms the B^{-1} exponent for the simulated dispatchers. So the functional form is not coming out of thin air, but it also isn't derived from dispatcher dynamics, and the paper's own conclusion admits it may not hold for other dispatchers. That limits the claim of universality across dispatching criteria. The main text says the scaling is 'insensitive' to dispatcher choice, while Fig. S4 shows different dispatchers give different finite-size universal curves that only asymptotically agree. The abstract overreaches.\n\nSecond, the collapse is partly built in: B1/2 is defined as the half-efficiency fleet size and then used to rescale. This would be circular if B1/2 were only fitted, but the mean-field predictions in Table S3 provide an independent anchor, so the collapse is not vacuous. Third, the efficiency is evaluated at finite x (7.5, 2.5) rather than the formal x->inf limit; the convergence is shown only qualitatively, and the main collapse figures have no error bars. A referee should ask for error bars and a convergence check.\n\nWho is this for? People modeling mobility-on-demand with statistical physics tools, and anyone estimating fleet sizes for new ride-sharing regions. It deserves a serious referee; with revisions to temper the dispatcher-universality wording and add the missing error analysis, it could be a solid publication.","headline":"Useful scaling result with a clean mean-field core, but the dispatcher-universality claim is overbroad and the finite-x collapse needs error bars.","tokens_in":17924,"tokens_out":3656,"would_cite":true,"duration_ms":35548,"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":"Ride-sharing efficiency across street networks collapses onto one universal curve set by two parameters.","keywords":["ride-sharing","ride pooling","universal scaling law","street networks","complex networks","on-demand mobility","efficiency measure","fleet size"],"falsifier":"Fix a network and request distribution, hold the normalized load $x$ constant, and measure the mean pickup waiting time $\\langle t_w\\rangle$ and the efficiency $E$ over a wide range of fleet sizes $B$. If $\\langle t_w\\rangle$ scales as $B^{-\\alpha}$ with $\\alpha\\neq 1$, or if the residuals from fitting $E=E_{\\max}B/(B+B_{1/2})$ grow systematically with $B$, the claimed universal collapse fails.","tokens_in":16820,"feed_emoji":"🚌","tokens_out":10438,"duration_ms":86695,"temperature":0.7,"pith_summary":"This paper proposes a generic, system-intrinsic measure of ride-sharing efficiency and claims that, across very different street network topologies, all efficiency curves collapse onto one universal curve. Under this scaling law, every service region is summarized by just two numbers: the maximum possible efficiency $E_{\\max}$ and the half-efficiency fleet size $B_{1/2}$, the number of buses needed to reach half of that maximum. The claim matters because it would let planners predict how efficient a ride-sharing service will be in a city, island, or rural area they have never serviced, and how many vehicles are required, without running a full simulation of every local detail. The same collapse is reported to hold for model networks and empirical street networks, and to be insensitive to changes in request distributions and dispatching algorithms.","feed_headline":"Ride-sharing efficiency collapses onto one universal curve","feed_subtitle":"Every street network reduces to two numbers: peak efficiency and half-efficiency fleet size.","key_machinery":"The central object is the efficiency $E$, built from the average number of scheduled customers per bus $\\langle C\\rangle$ and the normalized load $x$. The identity that carries the argument is $\\langle C\\rangle=\\frac{vx}{\\langle l\\rangle}(\\langle t_d\\rangle+\\langle t_w\\rangle)$, which ties efficiency directly to the average driving time $\\langle t_d\\rangle$ and waiting time $\\langle t_w\\rangle$. In the perfect-service limit the assumptions $\\langle t_d\\rangle\\sim \\langle l\\rangle/v\\propto B^0$ and $\\langle t_w\\rangle\\sim \\gamma(\\langle l\\rangle/v)B^{-1}\\propto B^{-1}$ convert this identity into $\\langle C\\rangle\\sim x(1+\\gamma/B)$, giving $E=E_{\\max}/(1+B_{1/2}/B)$ with $B_{1/2}=\\gamma$. A secondary object is the topological distinctness ratio $\\ell=l_{\\mathrm{tot}}/\\langle l\\rangle$, which quantifies how strongly shortest paths overlap and is shown to control the size of $B_{1/2}$.","core_discovery":"The central claim is that on-demand ride-sharing efficiency obeys a universal scaling law. Defining efficiency as $E=\\lim_{x\\to\\infty}(\\langle C\\rangle/x)^{-1}$, where $\\langle C\\rangle$ is the average number of scheduled customers per bus and $x=\\langle l\\rangle \\lambda/(vB)$ is the normalized request rate, the paper reports by extensive simulation that all efficiency curves collapse to $E=E_{\\max} f(B/B_{1/2})$, with a universal function $f(z)=1/(1+z^{-1})$ in the large-$z$ limit. $B_{1/2}$ is the fleet size at which the system reaches half of its maximum efficiency $E_{\\max}$; it absorbs the effect of the network topology and the request distribution, while $E_{\\max}$ is essentially set by the dispatching algorithm. The collapse is demonstrated on model networks spanning a minimal graph, rings, complete graphs, toroidal lattices, random geometric networks, a Cayley tree, and a star, and on empirical street networks from cities, islands, and rural areas.","pith_inferences":["Editorial inference: if the collapse holds, fleet sizing becomes a two-measurement problem: estimate $E_{\\max}$ and $B_{1/2}$ from a short pilot run in a new region, then read the whole efficiency curve off the universal function.","Editorial inference: the topological interpretation of $B_{1/2}$ could be sharpened by testing whether other graph metrics, such as path-overlap entropy or betweenness concentration, predict $B_{1/2}$ better than $\\ell=l_{\\mathrm{tot}}/\\langle l\\rangle$ on a wider set of empirical networks.","Editorial inference: the microscopic prediction hidden in the derivation is that pickup waiting time in a real fleet should halve when fleet size doubles; a field experiment varying only $B$ under fixed demand would test this independently of the efficiency collapse.","Editorial inference: if request distributions only renormalize $B_{1/2}$, then time-varying demand could be represented by a time-dependent $B_{1/2}(t)$ on the same universal curve, which would let operators track rush-hour efficiency with a single number."],"forward_implications":["Each street network is summarized by two parameters, $E_{\\max}$ and $B_{1/2}$, so its entire efficiency curve $E=E_{\\max}B/(B+B_{1/2})$ can be drawn without measuring every load level.","The half-efficiency fleet size $B_{1/2}$ grows with the distinctness of shortest paths, meaning a planner can use network geometry alone to estimate how many buses a new region will need.","Changing the request distribution only reshapes the scaling factor through the average trip length and effective topology; it does not change the functional form of the efficiency curve.","A dispatcher algorithm sets $E_{\\max}$ and can change $B_{1/2}$, but as long as waiting time decays as $B^{-1}$ for large fleets, the universal large-fleet form of the curve persists."],"supporting_citations":[{"why":"Supplies the shareability-network framework for quantifying the potential of vehicle pooling across cities, which this paper extends from theoretical potential to operating efficiency.","marker":"[18]"},{"why":"Provides the dynamic trip-vehicle assignment approach used for dispatching high-capacity ride-sharing in the simulations.","marker":"[20]"},{"why":"Establishes the earlier scaling law of urban ride sharing that this paper refines into a universal efficiency function.","marker":"[28]"},{"why":"Supplies the street-network extraction tool used to build the empirical networks of cities, islands, and rural areas.","marker":"[30]"}],"fun_headline_variants":["All ride-sharing networks share one efficiency curve","One curve rules ride-sharing efficiency in every network","Topology never alters ride-sharing efficiency law","A single function fits ride-sharing efficiency everywhere","Universal curve describes ride-sharing efficiency regardless of topology"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that, for a large fleet, doubling the number of buses halves the average waiting time for a pickup; if a dispatcher or network produces a different waiting-time scaling, the universal efficiency curve changes.","fun_headline_variants_meta":{"raw":{"variants":["All ride-sharing networks share one efficiency curve","One curve rules ride-sharing efficiency in every network","Topology never alters ride-sharing efficiency law","A single function fits ride-sharing efficiency everywhere","Universal curve describes ride-sharing efficiency regardless of topology"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001078,"raw_usage":{"total_tokens":4491,"prompt_tokens":908,"completion_tokens":3583,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":524,"completion_tokens_details":{"reasoning_tokens":3519}},"tokens_in":524,"tokens_out":3583,"duration_ms":23794,"temperature":1.0,"reasoning_tokens":3519,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:00:47.819526+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Fix a network and request distribution, hold the normalized load $x$ constant, and measure the mean pickup waiting time $\\langle t_w\\rangle$ and the efficiency $E$ over a wide range of fleet sizes $B$. If $\\langle t_w\\rangle$ scales as $B^{-\\alpha}$ with $\\alpha\\neq 1$, or if the residuals from fitting $E=E_{\\max}B/(B+B_{1/2})$ grow systematically with $B$, the claimed universal collapse fails.","supporting_citations":[{"cited_title":"Strogatz, and Carlo Ratti","cited_arxiv_id":null,"evidence_quote":"Supplies the shareability-network framework for quantifying the potential of vehicle pooling across cities, which this paper extends from theoretical potential to operating efficiency."},{"cited_title":"On-demand high- capacity ride-sharing via dynamic trip-vehicle assign- ment","cited_arxiv_id":null,"evidence_quote":"Provides the dynamic trip-vehicle assignment approach used for dispatching high-capacity ride-sharing in the simulations."},{"cited_title":"Scal- ing law of urban ride sharing","cited_arxiv_id":null,"evidence_quote":"Establishes the earlier scaling law of urban ride sharing that this paper refines into a universal efficiency function."},{"cited_title":"Osmnx: New methods for acquiring, con- structing, analyzing, and visualizing complex street net- works","cited_arxiv_id":null,"evidence_quote":"Supplies the street-network extraction tool used to build the empirical networks of cities, islands, and rural areas."}],"review_version":1}