From: Wang, To: "Randal Ruchti"; Sudhir Malik; "[email protected]" Cc: Assamagan, Ketevi; [email protected] Subject: Physics Education Meeting 3.30.2021 - Summary and Action Items Date: Wednesday, March 31, 2021 8:57:00 AM Attachments: image001.png

Dear All,

Below are notes and action items from yesterday’s meeting. Please, let me know if any changes are needed.

Physics Education Meeting March 30, 2021 1:00 PM – 2:10 PM EDT Attended: Randal Ruchti, David DeMuth, Marge Bardeen, Kenneth Cecire, Sijbrand de Jong, Yining You, Sudhir Malik, Vetri Velan, Savannah Thais, Olivia Bitter, Daria Wang (ORISE)

Summary: - CEF7 on Societal and Environmental Impacts was introduced during the last CEF All Convener meeting. - Physics Education upcoming fortnightly meetings are: April 13, 27, May 11, 25. Beginning June 1, intention would be to resume weekly meetings at this usual time. Can start with weekly meetings earlier if needed. - WP4_PE - Global Software Issues and HEP o LoI 21 - Open Science by and for HEP § HEP and Open Science The process of sharing scientific findings in journals has long been the mode of Open Science (OS) for a few centuries.

· Particle physics and Astronomy were among the first fields to embrace OS.

· Examples - open source software, open access to published work and open data for researchers and citizen scientists. o World Wide Web, Invenio, Indico, Inspire, Zenodo, OSG (Open Science Grid)are examples of some of the tools provided by the domain of particle physics to the scientific community and beyond. o CERN provides access to open source hardware, open publishing and open data via CERN Open Hardware License, Sponsoring Consortium for Open Access Publishing in Particle Physics (SCOAP3)and Open Data Portal for the LHC experiments. § Benefits

· Open Science empowers the ability to collaborate, question and contribute to sustainable scientific knowledge and process, and accelerate future discoveries. · Open data from LHC is used openly by educators, students, and self- learners to use and reuse for teaching, learning, and research.

· OS allows research to maximize its impact by reaching a wider audience of participation and sharing the findings, seeking collaborations without borders, and expanding funding and career advancement opportunities.

· OS is also greatly enhancing the quality and opportunities for education by making it more affordable and equitable.

· The strength of the particle physics HEP community lies in its openness, diversity and access to some of the best creative talent worldwide.

· With the advent of tools to operate OS, we are poised for an even greater and more significant role and in a more inclusive way. o For example § CERN open data is being used for public outreach, scientific work, Machine Learning and much more. § The software training in the HEP community is transforming the preparation of our next generation of problem solvers by reaching students beyond HEP. § Challenges

· On the other hand the particle physics community has to be aware and to make the outside world equally aware that there are practical limits, e.g. to making all raw data public.

· Besides problems of training the public at large to analyze these data, our computing infrastructure would not be able to support many others to access all our raw data. Our community should define, with good arguments, what should be the scope of making our data and resources publicly available. § Goals

· The goal of this LOI is to strengthen existing successful methodologies, many of which have been created and developed by the US HEP Community, and explore new ways of translating HEP resources, talent and data into Open Science.

· We can advance this powerful framework to create more inclusiveness by Open Software Training and Open Physics teaching that will propel our field and the general public into the next phase of discoveries. § Openness instills unimagined scale, expands engagement, facilitates diverse opinion, and allegedly prompts discoveries by amateurs, citizen scientists, and professionals. Unencumbered access to large data repositories alone are monolithic and overwhelming for a majority, proofs of concept, training, and access to a cadre of mentors needed to sustain progress.

· Action items: o Standardize software training, using a common open framework to develop and administer learning modules. o Inventory tried and trued open science methodologies. o Gauge, embrace, monitor and track open science related discoveries.

· Crosscutting: OpenStax development of introductory modern particle physics textbook.

· Related LOI’s: #2, #7. o LoI 30 - Supporting Research at the Intersection of Physics and Machine Learning § Motivation

· Machine Learning (ML) is becoming an integral part of physics research: many critical HEP algorithms for triggering, reconstruction, and analysis rely on ML

· Data-processing power of ML algorithms and these methods have quickly been adopted by physicists to address the unique timing, memory, and latency constraints of HEP experiments

· Unique constraints of physics experiments and the ability to exploit symmetries inherent in physics data have made the field of physics- informed ML a vibrant sub-field of Computer Science (CS) research

· Entire conferences and summer schools dedicated to this intersection § Challenges

· However, despite the relevance and importance of this research, pursuing a career at the intersection of these fields remains tenuous and undefined endeavor.

· How we can better support academic and career development at the intersection of physics and ML and ensure we can continue to benefit from and contribute to state-of-the-art ML.

· There are very few physics department supported courses on ML and interested students are often forced to convince CS departments to allow them into their graduate courses

· Furthermore, at many universities collaborative interactions between Physics and CS departments are very limited, which in turn limits the opportunities for physicists to engage with cutting-edge ML research and develop partnerships with industry researchers.

· As a physics graduate student at a US institute it is essentially impossible to focus one’s thesis research on ML as most departments do not acknowledge this as ‘physics work’ (despite the fact that these algorithms are in many cases critical to the functioning of physics experiments and data processing).

· Perhaps most egregiously, although there is now some support for this work at the postdoc or research staff level (see IRIS-HEP or the new IAIFI early-career researchers are often actively advised against applying to these positions as they constitute ‘career death sentences

· It is clear that these issues must be addressed in order to best support and advance our field, research, and researchers. § Solution

· An important first step is to work towards a cultural shift around the perception of what it means to ‘be a physicist’ and what work is considered as contributions to physics.

· We must value technical and software contributions in the same manner we value analysis work and foster a supportive and educational environment for graduate students and early-career researchers interested in ML and physics.

· It benefits no one to discourage individuals from pursuing work in this vibrant and promising field

· This necessitates developing cross-department courses, workshops, and research collaborations.

· Graduate programs should approach technical and computing skills with the same rigor they do traditional physics courses.

· Departments and advisors should be encouraged to allow students and postdocs to work across disciplines and explore opportunities for industrial partnerships or internships. This process should also include a reevaluation of requirements for physics graduate theses.

· The research landscapes of both physics and ML are shifting, and by adjusting to these changes we can ensure a vibrant future for both data-intensive physics and physics-informed ML

· Due to the lack of community statistics on this topic, the evidence presented here is primarily anecdotal although is a good example of how such a course might work NYU Physics collaborations with the Center for Data Science is an excellent example of cross-department collaboration § High dimensionality and large volumes of data are typical to HEP. Automating complex decision making using multivariate analysis utilizes networked and open science systems, requiring formal training and experience. Hit reconstruction, jet and track finding, particle identification, and event classification are prominent machine learning applications. Matriculating undergraduates who are adept across disciplines would kindle interest and study at the graduate level and simultaneously prepare some for computing jobs at national laboratories.

· Action items: o Physics undergraduates seek certifications in statistics, computer science, and software engineering. o Explore opportunities in undergraduate research. o Establish integrative physics, mathematics, and software engineering paradigms.

· Crosscutting: This LOI has overtones with the Computational Frontier.

· Related LOI’s: #12, #14, #20 o Abstract - Global Software Issues and HEP § ABSTRACT: The field of HEP leads in conducting open science. It enhances the ability to collaborate, question and contribute to sustainable scientific knowledge and process, and accelerate future discoveries. HEP has provided open source software, open access to published work and open data for researchers and citizen scientists. Several open collaborative tools have been developed for the scientific community and beyond: the World Wide Web, Invenio, Indico, Inspire, Zenodo, OSG (Open Science Grid), to name a few. There is also an increasing synergy with other fields like Computer Science where we benefit from recent progress in ML and utilized it in many critical HEP algorithms for triggering, reconstruction, and analysis. HEP open data is being used for public outreach, scientific work, Machine Learning and much more. The software training in the HEP community is transforming the preparation of our next generation of problem solvers by reaching students beyond HEP. While Open Science enables broader inclusiveness there are challenges to it in terms of providing support to non-HEP community to operate the open tools. There are also challenges to build in-house talent in software tools like Machine Learning as these increase in complexity and need a dedicated time investment. The current mindset that this investment is not “physics” work. The perception of what it means to ‘be a physicist’ and what work is considered as contributions to physics must expand. This WP explores these issues and suggested solutions. - Open discussion: o We don’t have serious career trajectories for students that are interested. Conflicting signaling in the field about what matters. Lack of meaningful career opportunities. This is happening in the field and we can try to make changes. o https://iaifi.org/ o https://iris-hep.org/ o A lot of people are encouraged to do hardware work because it makes you more competitive. Hardware is considered more physics than software. However, even if you do hardware all your life, it is hard to get faculty position. o What can we do through Snowmass to help our colleagues understand what Particle Physics requires in skills and talents and in terms of bringing them in? That’s big picture idea about what we can challenge the field to do. o Service tasks in software are a problem area. Software is considered as a tool. But there is more potential. Software is seen more and more as part of a detector. If detector is body, then software is brain. o Google and Facebook are better in data analysis (given their research teams, software etc.) than we are. We can’t just go out there and educate the world. It is important to strengthen those relationships. It’s good for our students to do internships at these companies. There is also interest from these organizations. Example of internship: Lamat Summer Research Program in Computational Astrophysics (UC). We should look into the internship effort more. It has been mostly about hardware. But partnerships in software would be good too. Support from supervisor is necessary to allow an internship to happen. 6 months’ time in their head is usually too much. But it is essential. o All of the emphasis is done on training PhDs – how many PhDs and how many faculty positions are going be there 10 years from now? If we can inspire undergrads with machine learning, if we want to grow the field and bring underrepresented groups, it’s hard to bring them in as particle physicists, but it is easier to get these people in in other ways. This WP might help with that. o From the San Francisco Declaration on Research Assessment, start 2nd paragraph: “The outputs from scientific research are many and varied, including: research articles reporting new knowledge, data, reagents, and software; intellectual property; and highly trained young scientists.” Conclusion: E.g. reconstruction software should be considered as a valuable output and used as such when assessing individuals when hiring and promoting. o Conferences reward young people who contribute to hardware. But there is no encouragement for those dedicated to software. Need to create something recognizable for software people.

Action items: - Next meeting on April 13, 2021: discussion on WP3 (Yining and Vetri) - undergrad and grad education.

Thank you,

Daria Wang, CMP, DES Event Project Planner Scientific and Technical Resource Integration (STRI)

865-839-7754 (cell) [email protected] Discover ORISE: Web | Facebook | | YouTube