1. The complete title of one (or more) paper(s) published in the open literature describing the work that the author claims describes a human-competitive result. Learning Emergency Medical Dispatch Policies via Genetic Programming Genetic and Evolutionary Computation Conference, 2023 Main contributions: 1. A modular GPHH learning framework to learn the five key decisions of Emergency Medical Dispatch (EMD). 2. A comprehensive statistical anaylsis, concluding we consistently improve on rules manually designed by experts by >50%. 3. A novel and publically accessible EMD dataset. 2. The name, complete physical mailing address, e-mail address, and phone number of EACH author of EACH paper(s). Authors: a) Jordan MacLachlan jordan.maclachlan@vuw.ac.nz +64 27 5995 885 b) Yi Mei yi.mei@vuw.ac.nz +64 4 886 5331 c) Fangfang Zhang fangfang.zhang@vuw.ac.nz +64 21 1145 371 d) Mengjie Zhang mengjie.zhang@vuw.ac.nz +64 4 4635 654 e) Jessica Signal jessica.signal@wfa.org.nz +64 4 499 9909 Address for authors (a) through (d): School of Engineering and Computer Science Room CO358, Cotton Building Victoria University of Wellington Gate 7, Kelburn Parade Wellington New Zealand Address for author (e): Wellington Free Ambulance 19 Davis Street Pipitea Wellington New Zealand 3. The name of the corresponding author (i.e., the author to whom notices will be sent concerning the competition). (a) Jordan MacLachlan jordan.maclachlan@vuw.ac.nz +64 27 5995 885 4. The abstract of the paper(s). Of great value to modern municipalities is the task of emergency medical response in the community. Resource allocation is vital to ensure minimal response times, which we may perform via human experts or automate by maximising ambulance coverage. To combat black-box modelling, we propose a modularised Genetic Programming Hyper Heuristic framework to learn the five key decisions of Emergency Medical Dispatch (EMD) within a reactive decision-making process. We minimise the representational distance between our work and reality by working with our local ambulance service to design a set of heuristics approximating their current decision-making processes and a set of synthetic datasets influenced by existing patterns in practice. Through our modularised framework, we learn each decision independently to identify those most valuable to EMD and learn all five decisions simultaneously, improving performance by 69% on the largest novel dataset. We analyse the decision-making logic behind several learned rules to further improve our understanding of EMD. For example, we find that emergency urgency is not necessarily considered when dispatching idle ambulances in favour of maximising fleet availability. 5. A list containing one or more of the eight letters (A, B, C, D, E, F, G, or H) that correspond to the criteria (see above) that the author claims that the work satisfies. (D, E, F, G) 6. A statement stating why the result satisfies the criteria that the contestant claims (see examples of statements of human-competitiveness as a guide to aid in constructing this part of the submission). re: (D): "The result is publishable in its own right as a new scientific result independent of the fact that the result was mechanically created." We worked closely with Wellington Free Ambulance (WFA) to design a set of heuristics that approximate the real world behaviour of their dispatch professionals. In our simulated environment, our method reduces the average weighted response time by up to 69%. If WFA were to achieve such performance improvements in practice, via an algorithmic techniques or otherwise, they would radically improve the lives of the populous they serve. Further, WFA would disseminate such knowledge via publications and conferences in their professional field. Undoubtedly, our technique offers promise in practice, and when we do such real-world trials, the results will be publishable. This paper achieved an average reviewer recommendation score of 4.3, with some of their feedback highlights as follows: - "The article is well written, the problem is well defined, the proposed algorithm is described by using pseudocodes, and the results are supported by statistical tests." - "The problem and the proposed solutions approach are engaging and fit in the [scope] of the conference and the track of real-world application." - "The paper is well-written and well-organised." re: (E): "The result is equal to or better than the most recent human-created solution to a long-standing problem for which there has been a succession of increasingly better human-created solutions." Civilian ambulances have existed since the 1800s; emergencies in the community, since long prior. Ever since, service providers have sought methods to optimise resource allocation. Only in recent years have automated techniques become available. Human dispatchers (i.e. domain experts) have long been favoured in industry due to their ability to audit fatal dispatch decisions. Pre-existing manually designed dispatch rules fail to account for the complexity of the real world, and perform poorly. In this work, we propose a new suite of manually designed dispatch rules that approximate WFA dispatch behaviour, explicitly co-designed with expert dispatchers at WFA. Further, our Genetic Programming Hyper Heuristic (GPHH) technique offers a path toward interpretable machine-learned dispatch policies that significantly outperform this manual suite. re: (F): "The result is equal to or better than a result that was considered an achievement in its field at the time it was first discovered." Alongside WFA, we manually co-designed the dispatch rule suite to approximate expert human decision making as existing rules in the literature were too naïve for use in practice (quoting industry). To illustrate this, WFA has manual rules implemented in their existing dispatch platform, yet defer to human decisions >95% of the time. From internal tests, we know our co-designed manual suite vastly outperforms these preexisting rules, and our GPHH technique vastly outperforms our manual suite. re: (G): "The result solves a problem of indisputable difficulty in its field." EMD is particularly difficult due to a) its inherent dynamism and uncertainty, b) ambulance collaboration requirements, c) interpretability requirements, and d) the sheer risk to life. EMD adds uncertainty, dynamism, collaboration, interpretability, and urgency requirements to other widely studied combinatorial optimisation problems such as vehicle/arc routing problems, which are already NP-Hard. 7. A full citation of the paper (that is, author names; title, publication date; name of journal, conference, or book in which article appeared; name of editors, if applicable, of the journal or edited book; publisher name; publisher city; page numbers, if applicable). J. MacLachlan, Y. Mei, F. Zhang, M. Zhang, and J. Signal, “Learning Emergency Medical Dispatch Policies via Genetic Programming,” in the Genetic and Evolutionary Computation Conference, 2023. 8. A statement either that "any prize money, if any, is to be divided equally among the co-authors" OR a specific percentage breakdown as to how the prize money, if any, is to be divided among the co-authors. All prize money, if any, will be allocated to the primary author to support his PhD studies. However, he will donate 20% of such winnings to Wellington Free Ambulance to acknowledge their contributions. 9. A statement stating why the authors expect that their entry would be the "best". This paper uses GPHH for solving an extremely important real-world problem. Improving ambulance response times translates directly to saved lives; mortality increases by ~24% for each minute a cardiac patient waits for an ambulance. Not only do we propose a manual rule suite codesigned with domain experts that approximates 2023 industry behaviour, but an algorithmic technique that significantly improves weighted response time performance. Further, alongside industry, we design and publish the first public-access EMD dataset. For each of the 44 graphs provided, we provide 1000 EMD instances (each a 24-hour period) upon which future scholars can learn. We aim to do good in the world by advancing humanity's technological capacity. Should that not be our core objective, as computer scientists? 10. An indication of the general type of genetic or evolutionary computation used, such as GA (genetic algorithms), GP (genetic programming), ES (evolution strategies), EP (evolutionary programming), LCS (learning classifier systems), GI (genetic improvement), GE (grammatical evolution), GEP (gene expression programming), DE (differential evolution), etc. Genetic Programming (Hyper Heuristic). 11. The date of publication of each paper. If the date of publication is not on or before the deadline for submission, but instead, the paper has been unconditionally accepted for publication and is “in press” by the deadline for this competition, the entry must include a copy of the documentation establishing that the paper meets the "in press" requirement. In-press for GECCO 2023. See below accepatance email: Dear Yi Mei, Your paper (pap396s2) Learning Emergency Medical Dispatch Policies Via Genetic Programming has been accepted as a *full paper* (page limit: 8 pages + 2 pages with references) for GECCO 2023. Congratulations! Reviews are now available in the submission website at https://ssl.linklings.net/conferences/gecco. Please confirm or decline acceptance in the link above. Acceptance is subject to the condition that you consider the comments of the reviewers when preparing the camera-ready version of your manuscript. After confirming acceptance, you will receive a separate email with instructions and information about the copyright process. You need to complete the copyright form to get the copyright notice with the final DOI you have to include in the camera-ready version of the manuscript. Your camera-ready version must be submitted via the submission system (https://ssl.linklings.net/conferences/gecco/) by Thursday, April 20th, 2023. Please, prepare your camera-ready version following the instructions and templates at https://gecco-2023.sigevo.org/Paper-Submission-Instructions (see the Camera-Ready Instructions). Upload preliminary versions of your manuscript prior to the deadline to check if there is any problem or formatting issue (e.g., inappropriate or not embedded fonts). Please note that May 10th, 2023 is the deadline for authors of accepted papers/posters to register. At least one author for each accepted paper/poster must be registered by then, and at least one author must present the paper at the conference. The registration site will open soon at https://gecco-2023.sigevo.org/Registration Also, note that this work should not be submitted to any GECCO 2023 Workshop. Thank you for submitting your work to GECCO 2023. I look forward to seeing you in Lisbon! Kind regards, Luís Paquete EiC GECCO 2023 ------------- And below confirmation of submitting our camera-ready verison: Dear Jordan MacLachlan, The camera-ready material of your submission, "Learning Emergency Medical Dispatch Policies Via Genetic Programming" (pap396s3), to Genetic and Evolutionary Computation Conference (GECCO-2023), has been received. You can update this submission until the deadline: 11:59pm (UTC-12) Apr 20, 2023. The one and only URL you need for all things pertaining to your submission is: https://ssl.linklings.net/conferences/gecco/ You can and should verify that the information you submitted was properly received by signing into the submission website. If you have forgotten your password, use the 'forgot password?' link on the submission website front page. After signing into the website you may also modify or withdraw your submission (until submissions close). Thank you for your submission! Luís Paquete, Editor-in-Chief paquete@dei.uc.pt https://gecco-2023.sigevo.org ------------- Thank you for your consideration!