{"id":"b886803c-6fc6-4db4-86df-f3677a59eff0","arxiv_id":"2501.09479","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":3.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"A qualitative field survey in Kathmandu finds that residents across three neighborhoods consistently want faster, cheaper, more reliable internet, which the paper argues is the key enabler for AI-enabled city services in emerging economies.","lead":"This paper surveys internet users in three Kathmandu neighborhoods and combines the results with a review of how connectivity affects AI-enabled city services in emerging economies. It reports that residents across income levels want faster, cheaper, more reliable internet, and that affordability and infrastructure gaps block AI applications.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 'proper pricing prevents congestion' conclusion in §12.5 rests on an unvalidated simulation with only two assumed parameters (10x price ratio, 10 Mbps threshold) and no sensitivity analysis; if the model is not calibrated to actual Kathmandu prices and traffic, the central policy claim is…","rationale":"The reader's verdict is CONDITIONAL, and I agree no outright rejection is warranted; however, the most fragile part of the central claim is not only the survey's representativeness but the causal inference in Section 12.5. The paper claims pricing is key to avoiding congestion, yet the simulation supporting that claim is uncalibrated and has only two free parameters. The authors report a Wi-Fi subscription price of NPR 1430/month but never give the corresponding cellular price or user switching data. The 10 Mbps threshold is asserted without justification. A simple sensitivity check with real price ratios could either rescue or refute the conclusion. I therefore recommend keeping the CONDITIONAL verdict but attaching an explicit condition: release the survey instrument/raw data and re-run the simulation with calibrated parameters and sensitivity analysis. This concern is complementary to the reader's representativeness concern, hence 'partial'.","tokens_in":15453,"tokens_out":6982,"duration_ms":67713,"concrete_test":"Obtain the actual Wi-Fi and cellular data prices for the three surveyed areas and the raw survey dataset. Re-specify the §12.5 simulation with explicit demand and switching equations, then run it with the actual price ratio (replacing the assumed 10x) and vary the congestion threshold from 5 to 20 Mbps. If the Wi-Fi congestion episodes in Figure 8 disappear under realistic ratios or reasonable thresholds, or if the model's equilibrium does not reproduce the observed oscillation, the 'proper pricing is key' claim is an artifact of the assumed parameters. As an independent check, recompute Figures 1 and 2 from the raw data and confirm the percentages sum to 100 if the questions were single-choice; a failure here indicates the survey data cannot support any quantitative claim.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 12.5 presents a techno-economic simulation to support the claim that 'proper pricing of internet services is key to avoid network congestion.' The simulation is described only by two assumptions: cellular data priced at ten times Wi-Fi, and congestion defined as average bandwidth per user below 10 Mbps. No model equations, traffic demand distribution, user decision rule, or network capacity parameters are given, and no sensitivity analysis is reported. For the causal claim to hold, user choices must actually respond to the price differential, the 10x ratio must reflect the real market in Kathmandu, and the 10 Mbps threshold must be the appropriate QoS measure. The paper does report Wi-Fi cost (NPR 1430/month in Bansighat) but no comparable cellular price, no measured user switching behavior, and no objective bandwidth data. The statement that the phenomenon 'was observed in the deployed network' is anecdotal. Without calibration and sensitivity, the conclusion that pricing—rather than capacity, contention, or infrastructure quality—is the key lever is not established. The survey side is also underdocumented: no sample size, questionnaire, or response rate, and Figures 1-2 percentages do not sum to 100, which further undermines the empirical basis.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper argues that the success of AI-enabled urban applications in emerging economies depends on reliable, affordable telecommunications connectivity. After a broad review of AI in urban governance, healthcare, sustainability, labor, and economics, it presents a survey conducted in three Kathmandu neighborhoods (Asan, Bansighat, Kusunti) and claims that residents aspire to high-speed, reliable, low-cost internet. It also introduces a techno-economic simulation in Section 12.5 intended to show that cellular data priced much higher than Wi-Fi leads to Wi-Fi congestion, and it concludes that proper pricing of internet services is key to avoiding network congestion. The paper's central message is that infrastructure investment, affordability, and QoS are prerequisites for AI-enabled cities in emerging economies.","tokens_in":15686,"tokens_out":5452,"duration_ms":55566,"significance":"If the empirical claims were fully supported, the paper would make a useful contribution to the digital-divide and smart-city literature by grounding connectivity barriers in field data from an under-studied city. The qualitative descriptions of user experiences, the reported cost differential between Wi-Fi and cellular data, and the identified congestion and reliability complaints are plausible and policy-relevant. The paper also usefully connects a set of AI-for-urban-life applications to the often-overlooked network infrastructure constraint. However, the paper's empirical core is not currently verifiable: the survey methodology is undocumented, the demographic figures contain internal inconsistencies, and the Section 12.5 simulation is not reproducible. The paper does not ship machine-checked proofs, code, or a full instrument, so its value rests entirely on the adequacy of the reported survey and simulation, which are at present insufficiently described.","major_comments":[{"comment":"The paper does not report the survey sample size, sampling method, respondent selection criteria, questionnaire, response rate, or analysis procedure. The only methodological information is that three Kathmandu locations were chosen and that respondents were asked about their internet experiences. Without these details, the reader cannot assess selection bias, question wording effects, or whether the responses support the strong generalization in Section 13 that 'the people of emerging economies have high aspirations' for high-speed internet. Please add a complete methods subsection, including the instrument, the number of respondents per site, and the recruitment procedure.","section":"Section 12 and Section 13"},{"comment":"The reported demographic distributions are internally inconsistent: the percentages in Figure 1 (45+39+14+20) sum to 118%, and those in Figure 2 (8+70+3+17+10) sum to 108%. These sums cannot both be valid percentage distributions, so the demographic context of the survey is unclear. Please correct the values, state the exact category definitions, and explain how rounding was handled.","section":"Section 12, Figures 1 and 2"},{"comment":"The conclusion that 'proper pricing of internet services is key to avoid network congestion' rests on a simulation that is not described in sufficient detail to be checked or reproduced. The text gives only two assumptions (cellular data priced at ten times Wi-Fi, and congestion defined as average bandwidth per user below 10 Mbps) and reports no model equations, demand distribution, user switching rule, network capacity parameters, calibration to Kathmandu prices or traffic, or sensitivity analysis. The statement that the phenomenon 'was observed in the deployed network' is anecdotal and does not substitute for validation. Please either provide the full model with calibration and sensitivity results, or reframe the conclusion as a hypothesis with explicit limitations.","section":"Section 12.5"},{"comment":"The paper generalizes from three purposively selected neighborhoods in a single city to 'emerging economies' in the title, abstract, and discussion. There is no evidence that Asan, Bansighat, and Kusunti are representative of the diversity of urban connectivity conditions across emerging economies, and the survey captures perceived QoS rather than objective network measurements. Please narrow the scope of the claims to Kathmandu or provide a reasoned sampling justification and external validity evidence.","section":"Abstract, Section 11, and Section 13"}],"minor_comments":[{"comment":"The keyword 'emerging economics' should be 'emerging economies', and Section 1 contains the phrase 'In all workplace' which should be 'In the workplace'.","section":"Keywords and Section 1"},{"comment":"The sentence 'proper pricing is internet services is key' contains a typo and should read 'proper pricing of internet services is key'.","section":"Section 12.5"},{"comment":"The word clouds and word-frequency plots would benefit from a description of the text-processing steps (stopword removal, stemming, part-of-speech filtering) and the number of responses used to generate them.","section":"Figures 5 and 6"},{"comment":"Reference [38] is a book review of Zuboff's 'The Age of Surveillance Capitalism'; as cited, it does not directly support the claim about legal frameworks for AI in Section 6. Please replace it with a primary source on AI regulation or clarify the connection.","section":"Reference [38]"},{"comment":"Several statements such as 'One-third of them use mobile banking applications' and 'On average, users are paying NPR 1430/month' would be more useful if accompanied by the relevant sample sizes and the number of respondents making each statement.","section":"Section 12.1 and 12.2"}],"recommendation":"major_revision","confidential_remarks":"The manuscript reads more like a position paper or project report than a full research article. The core empirical claims depend on data and methods that are not currently in the paper, and the Section 12.5 simulation is presented only as an illustration. If the authors can supply a proper methods section, correct the demographic figures, and either validate or substantially qualify the simulation, a revised version could be viable for this venue. The fit with cs.CY is appropriate given the socio-technical focus on connectivity and AI in cities."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe paper's only real asset is the descriptive survey from three Kathmandu neighborhoods (Asan, Bansighat, Kusunti). That data is new, and the qualitative narrative—people want affordable, reliable broadband and find cellular data expensive—is credible and consistent with the digital divide literature. Everything else is a restatement of known AI-for-cities applications and the obvious point that connectivity is a prerequisite.\n\nThe soft spots are not subtle. First, the survey is reported without sample size, sampling method, instrument, or response rate. We are told the locations were chosen to represent commercial, low-income, and high-income areas, but we cannot assess whether respondents are representative or whether the complaints track actual network performance. Second, Figures 1 and 2 have percentage bars that sum to 118% and 108% respectively; that alone requires correction before any number can be trusted. Third, the §12.5 \"techno-economic simulation\" is not a simulation in any usable sense: it rests on two assumptions (cellular priced at 10x Wi-Fi, congestion below 10 Mbps per user) with no equations, demand model, or sensitivity analysis. The claim that \"proper pricing ... is key to avoid network congestion\" is asserted from this toy. The observation that the phenomenon was seen in the deployed network is anecdotal.\n\nI agree with the stress-test note that the pricing conclusion is not established. That said, I don't think there is circularity beyond the usual framing—the survey was designed around connectivity, so of course it finds connectivity matters. The central conclusion is mainstream, which is fine, but it was already mainstream before this paper.\n\nWho is this for? Someone looking for a single citation that 'emerging-economy residents want affordable broadband' might use it, but the empirical basis is too shaky to cite as data. I would not bring it to a reading group except as an example of underdocumented survey reporting. As a referee assignment, I would not send it out in its current form: the missing methodology and the internal inconsistencies are fixable, but the simulation is too thin to salvage without real calibration or dropping the claim. If the authors resubmit a short, honest data report with the actual survey instrument and raw aggregates, that would be worth a look.","headline":"A likeable but thin field survey: the Kathmandu data are new and plausible, yet the missing methodology and a toy simulation leave the paper short of a research contribution.","tokens_in":16221,"tokens_out":2508,"would_cite":false,"duration_ms":25867,"reading_group":"no","serious_thinker":"yes","would_accept_peer_review":false},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A Kathmandu field survey finds connectivity gaps, not lack of demand, are what block AI-enabled cities in emerging economies.","keywords":["artificial intelligence","smart cities","telecommunications infrastructure","internet affordability","Wi-Fi congestion","digital divide","emerging economies","Kathmandu survey"],"falsifier":"Run continuous speed tests in Asan, Bansighat, and Kusunti during the evening and weekend hours when residents reported slowdowns, and compare measured per-user bandwidth to the survey responses; if average bandwidth stays above 10 Mbps during those peak periods, the paper's congestion diagnosis would be directly contradicted. A complementary test would lower cellular data prices in one neighborhood while holding a matched comparison neighborhood unchanged and observe whether Wi-Fi congestion actually decreases, as the pricing mechanism predicts.","tokens_in":15250,"feed_emoji":"📶","tokens_out":9382,"duration_ms":83424,"temperature":0.7,"pith_summary":"Before AI-enabled urban services can succeed in emerging economies, the connectivity problem has to be solved first. This paper argues that a field survey of three Kathmandu neighborhoods — a commercial hub, a low-income settlement, and a high-income settlement — shows residents already use the internet heavily and uniformly want faster, cheaper, more reliable service. The authors claim the main barrier is not lack of demand but inadequate telecommunications infrastructure plus a price structure that pushes users onto congested shared Wi-Fi, and they simulate how that congestion develops when cellular data costs ten times as much as Wi-Fi. Because more than three billion people are unserved or underserved and most of the world's population lives in emerging economies, the paper concludes that closing this gap is a precondition for AI applications in urban areas.","feed_headline":"Connectivity gaps block AI in emerging cities, Kathmandu survey shows","feed_subtitle":"Kathmandu residents want fast, affordable internet; without it, smart-city AI will stay out of reach for most.","key_machinery":"The empirical engine is a deliberately chosen field survey of three Kathmandu areas — Asan, a commercial hub; Bansighat, a low-income settlement; and Kusunti, a high-income area — capturing age, education, gender, occupation, internet usage, and complaints. The explanatory mechanism is a techno-economic simulation of Wi-Fi congestion that assumes cellular data costs ten times as much as Wi-Fi and declares congestion when average bandwidth per user drops below 10 Mbps; the simulation reproduces the alternating crowding and retreat the survey respondents described. This pricing mechanism is what carries the paper's claim that affordability, not just infrastructure, drives network quality.","core_discovery":"On the paper's own terms, the discovery is that affordable, reliable connectivity is the missing precondition for AI-enabled cities in emerging economies. The Kathmandu survey finds that in all three settings people use the internet for communication, entertainment, banking, learning, and business, and that they consistently report slowdowns, buffering, and high costs, with congestion worst in the evenings and on weekends. The accompanying simulation shows that when cellular data is priced at ten times Wi-Fi, users crowd onto shared Wi-Fi, average bandwidth per user falls below the 10 Mbps congestion threshold, and the cycle of crowding, retreat, and return repeats, matching the survey's complaints. The paper therefore concludes that proper pricing of internet services is key to avoiding congestion and that network investment, public or public-private, is needed to make AI-driven urban life a reality in emerging countries.","pith_inferences":["A direct test the paper does not report: measure objective throughput in the same neighborhoods at the same peak times and compare it with the survey's perceived slowdowns, since perception and measured bandwidth can diverge.","The pricing mechanism implies a testable elasticity: if cellular data prices drop in one neighborhood but not in a matched control, Wi-Fi congestion should fall measurably in the treated area, which would confirm the causal story.","The paper's call for free internet for essential services sits in tension with its own congestion simulation; a free tier would need prioritization or capacity guarantees, otherwise it may simply move the congestion problem.","The survey's three purposively selected sites make the generalization to all emerging economies a hypothesis rather than an established fact; a multi-city, random-sample survey with objective network data would be the natural next study."],"forward_implications":["If correct, the price gap between cellular data and shared Wi-Fi is an actionable lever: repricing cellular data could relieve congestion even before new infrastructure is built.","If correct, AI applications in urban governance, healthcare, finance, and public safety will remain unreliable in emerging-economy cities until last-mile networks and investment gaps are closed.","If correct, network slicing and spectrum sharing could protect essential services by giving them guaranteed quality even when consumer traffic overloads shared connections.","If correct, free or subsidized internet for essential services would help bridge the digital divide, but only if capacity and pricing are managed to avoid recreating the congestion loop.","If correct, the Kathmandu pattern should be found in other dense emerging-economy cities with similar income mixes, making it a general barrier rather than a local one."],"supporting_citations":[{"why":"Supplies the scale figure of more than three billion people without adequate internet access, framing the problem the paper addresses.","marker":"[14]"},{"why":"Supplies the 'Connecting the Unconnected' initiative that frames the survey project and its intervention goal.","marker":"[15]"},{"why":"Supports the premise that AI-for-cities applications are distributed and therefore depend on telecommunications networks.","marker":"[16]"},{"why":"Underpins the claim that modern 5G-and-beyond networks are the enabling infrastructure for distributed AI applications in cities.","marker":"[63]"},{"why":"Provides the network-slicing mechanism the paper proposes for guaranteeing quality of service to essential urban services.","marker":"[68]"},{"why":"Supplies a nomadic-node deployment case study in Nepal treated as a feasible path for connecting currently unserved areas.","marker":"[69]"}],"fun_headline_variants":["Affordable internet is the missing link for AI cities","Kathmandu survey: pricey data stalls smart-city AI","Without cheap bandwidth, AI urban services fail","High costs, congestion block AI in emerging cities","AI cities need reliable internet, Nepal case shows"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the three deliberately chosen Kathmandu neighborhoods, and residents' self-reported complaints about internet service, stand for urban connectivity problems across emerging economies; the paper provides no sampling frame, response rate, or independent network measurements to anchor that extrapolation.","fun_headline_variants_meta":{"raw":{"variants":["Affordable internet is the missing link for AI cities","Kathmandu survey: pricey data stalls smart-city AI","Without cheap bandwidth, AI urban services fail","High costs, congestion block AI in emerging cities","AI cities need reliable internet, Nepal case shows"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000236,"raw_usage":{"total_tokens":1507,"prompt_tokens":949,"completion_tokens":558,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":565,"completion_tokens_details":{"reasoning_tokens":484}},"tokens_in":565,"tokens_out":558,"duration_ms":6890,"temperature":1.0,"reasoning_tokens":484,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T19:57:57.611547+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run continuous speed tests in Asan, Bansighat, and Kusunti during the evening and weekend hours when residents reported slowdowns, and compare measured per-user bandwidth to the survey responses; if average bandwidth stays above 10 Mbps during those peak periods, the paper's congestion diagnosis would be directly contradicted. A complementary test would lower cellular data prices in one neighborhood while holding a matched comparison neighborhood unchanged and observe whether Wi-Fi congestion actually decreases, as the pricing mechanism predicts.","supporting_citations":[{"cited_title":"Broadband Infrastructure, Access and Use","cited_arxiv_id":null,"evidence_quote":"Supplies the scale figure of more than three billion people without adequate internet access, framing the problem the paper addresses."},{"cited_title":"Connecting the Unconnected","cited_arxiv_id":null,"evidence_quote":"Supplies the 'Connecting the Unconnected' initiative that frames the survey project and its intervention goal."},{"cited_title":"R., Singh, V","cited_arxiv_id":null,"evidence_quote":"Underpins the claim that modern 5G-and-beyond networks are the enabling infrastructure for distributed AI applications in cities."},{"cited_title":"Connecting the Unconnected: A DT Case Study of Nomadic Nodes Deployment in Nepal","cited_arxiv_id":"2411.09380","evidence_quote":"Supplies a nomadic-node deployment case study in Nepal treated as a feasible path for connecting currently unserved areas."}],"review_version":1}