Whether hailed as an arresting movie title (Kwan & Scheinert, 2022) or dismissed as an attention-grabbing phrase, “everything everywhere all at once” aptly encapsulates the anxiety and overwhelming deluge many of us feel amidst the onslaught of—for lack of a better phrase—AI-rrhea. Conversations during a recent HDSR awareness tour in Latin America reminded me vividly of two prevalent anxieties: FOMO (fear of missing out) and FOGO (fear of grave outcome). My dual titles as a professor of statistics and the editor-in-chief of a data science journal seemed to act as a magnet for a barrage of pressing questions. Everyone else seems to be trying AI on everything, everywhere, but where should I start? What should I do? Am I already too late? How will AI affect my job or life? Will I even have a job or life?
I share the anxiety of these inquirers—not because I personally fear missing out or worry about being replaced. I’m already overcommitted to the overcommitted feeling, and frankly, I would be thrilled if AI could take over most of what I do, aside from writing this editorial column. However, in my role as editor-in-chief, I do experience both FOMO and FOGO for HDSR. There are so many topics to cover, partnerships to forge, and initiatives to pursue. Where should we start, and how do we manage the workload with a three-staff editorial office, including me (with no teaching or research reduction)? As we navigate a world where rapidly evolving digital technologies are reshaping how we pursue, present, and publish knowledge advances and intellectual engagement, I wonder: Will we celebrate the 10th anniversary of HDSR with the same joy and optimism we felt at its 5th, if we simply continue with business as usual?
The desire to feature everything and be everywhere was, in fact, the driving force behind HDSR’s founding, as reflected in its mission tagline: everything data science and data science for everyone. Its vision tagline further amplifies this ambition: a telescopic, microscopic, and kaleidoscopic view of data science. Yet, neither tagline fully captures the immediacy or overwhelming intensity conveyed by ‘all at once.’ Indeed, one could argue that the vision tagline emphasizes careful contemplation and in-depth investigation—qualities that cannot be rushed or achieved under relentless pressure.
The ultimate challenge for HDSR, therefore, is this: How can it be everything, everywhere, all at once, while remaining the most reliable source for telescopic visions, microscopic investigations, and kaleidoscopic reflections in data science and AI? Is such a balance even possible?
This is not a rhetorical question meant to prompt a resounding ‘Yes, we can!’ nor is it a veiled admission of defeat before attempting. Instead, it is an invitation—to you, the readers of HDSR—to brainstorm with me as we explore the 16 articles in this issue 7.1, a collection that epitomizes HDSR’s ‘everything everywhere’ spirit. I warmly welcome your suggestions (please write to [email protected]) on strategies to help HDSR fulfill its mission and expand its vision in this rapidly evolving, all-at-once age of AI, regardless of your stance on AI itself.
The section title paraphrases the interview conducted by Hamit Hamutcu, co-editor of HDSR’s Active Industrial Learning column, titled “Navigating the Rapid Evolution of AI With Todd James” (James & Hamutcu, 2025). Todd James is the chief data and technology officer at 84.51˚, a customer-focused data science, analytics, and consulting organization serving as the in-house analytics arm for the Kroger Company, one of the largest grocery retailers in the United States. Positioned at the forefront of leveraging AI technologies to transform practice into data and data into consumer satisfaction, James’s insights and wisdom shed valuable light on the power and pitfalls of AI for decision-making in business and beyond. He also underscores the importance of fostering effective human-AI interaction, particularly in an era when developments are unfolding rapidly and disruptively.
This issue also features another interview conducted by Hamutcu for the Active Industrial Learning column: “In Conversation With Intuit Chief Data Officer Ashok Srivastava” (Srivastava & Hamutcu, 2025). Intuit, a leading financial software and technology company in the United States, provides tools for financial management, tax preparation, and business operations. The discussion focuses on Srivastava’s vision for AI, his work at Intuit, and how education supports the company’s mission to “power prosperity around the world.” As Hamutcu highlights, Srivastava’s direct involvement in developing curricula and training employees to understand cutting-edge AI, generative AI, machine learning, data science, and experimental design reflects his role as an adjunct professor at Stanford University.
As a university professor myself, I was particularly delighted to hear Srivastava’s emphasis on skill training. I look forward to personally thanking him on February 6 this year, when Intuit’s headquarters will host the inaugural “Conversations with HDSR.” Centered around the theme “Beyond the Agentic AI Hype,” this event will feature a fireside chat between Srivastava and Tom Davenport (former co-editor of the Active Industrial Learning column), as well as a panel discussion moderated by Hamutcu’s current co-editor, Miguel Paredes. I am deeply grateful to Hamutcu for spearheading the Conversations with HDSR initiative as part of our outreach efforts to improve data science and AI literacy across industry, business, and beyond. (If your organization is interested in hosting such conversations, please feel free to contact me at the editor-in-chief email address shared earlier.)
With these broad goals in mind, it is especially heartening to hear Todd James’s emphasis on education beyond the users of technologies. As James stated in his interview (James & Hamutcu, 2025), “When you start enabling decisions with increased data, with advanced analytics, that has a direct impact on people. It’s very important to educate them so that they can play a role.” Indeed, increasing public data literacy is at the core of the ‘data science for everyone’ aspect of HDSR’s mission. I am therefore particularly grateful to Hamutcu and his former co-editor, Katie Malone, for organizing a data literacy workshop cosponsored by the Harvard Data Science Initiative and the Institute for Experiential AI at Northeastern University in December 2023.1
The workshop led to a special theme, highlighted in Hamutcu’s (2025) editorial, “Introduction to Special Theme on Data Literacy at Scale,” which is being published in two parts across this Winter issue and the next Spring issue of HDSR. Without repeating the workshop summary or the introductions to the two featured articles in Hamutcu’s editorial, I wish to highlight that the titles alone—“Data Literacy in Industry: High Time to Focus on Operationalization Through Middle Managers” (Koloski et al., 2025) and “Building an Inclusive Data Literacy Community” (Hilty et al., 2025)—underscore the timeliness and pervasive need for data literacy in everything, everywhere.
Once again, this section title borrows from an article title in this issue, “Data Science for Central Banks and Supervisors: How to Make It Work, Actually” (Duijm & van Lelyveld, 2025). The article’s abstract begins diplomatically: “Public authorities, such as central banks and supervisory authorities, are not known for their ability to quickly adopt new techniques in a rapidly changing world.” Although the sectors differ, the broad challenges addressed in this article mirror those reflected in the two interviews with industry leaders—ranging from data quality to employee knowledge and to communication among stakeholders. Focusing on how to make things work in practice, the authors offer nine specific lessons, spanning topics such as data quality, data governance, organizational engagement, and setting long-term goals. These lessons carry wide-ranging implications, not just for practical applications of data science and AI but also for theoretical and methodological developments. For instance, lesson 5 states, “In many cases, a simple model is a great place to start” (boldface in the original). This advice should evoke a vivid mental image for those of us adept at summoning complex theories and methods: just because we can does not mean we should.
Moving from finance to health, the article by Zhicheng Guo, Cheng Ding, Duc Do, Amit Shah, Randall J. Lee, Xiao Hu, and Cynthia Rudin (2025), “Improving Atrial Fibrillation Detection Using a Shared Latent Space for ECG and PPG Signals,” documents a substantial team effort to train a deep neural network for more accurate predictions of atrial fibrillation using data from consumer wearables. For life-threatening medical conditions that affect large populations, even a small reduction in actual predictive error can translate into saving thousands of lives and significant resources. However, achieving such gains requires more than technical expertise in developing and tuning deep neural networks. Issues such as data quality and ascertainment bias can easily undermine potential benefits—or worse, turn methodological advances into actual harm. I therefore thank the authors for not only documenting their achievements but also candidly discussing the limitations of their data and the challenges they encountered, providing proper cautions to users and motivations for further improvements.
Indeed, with the rapid advancements in pattern-seeking AI and generative AI, it is well understood that the effectiveness of these technologies depends critically on the quality of the data they process. Ensuring reliable data at all scales has become more crucial than ever, especially since AI algorithms will continue producing results regardless of data quality—at least for now. However, collecting, processing, and maintaining reliable data sets is a complex and costly enterprise. These challenges are starkly highlighted in “The Nation’s Data at Risk: The First Annual Report on the Federal Statistical System” by Jonathan Auerbach, Claire McKay Bowen, Constance F. Citro, Steve Pierson, and Zachary H. Seeskin (2025). The challenges are increasingly intensified because of “declining survey response rates, shrinking budgets (in real terms), and increasing threats to data confidentiality that have prompted agencies to restrict more data—not to mention increasingly brazen attempts to exert undue influence on agency operations for political gain.”
Addressing these issues will require a tremendous collective effort. I deeply appreciate the painstaking work reported in the Auerbach (2025) article of this issue, done by a team of researchers from the American Statistical Association and George Mason University, and their invitation for comments and suggestions from HDSR readers. I strongly encourage concerned readers to reflect on the six guiding questions that framed the report’s examination of “the fundamental requirements for and challenges to an effective federal statistical system,” and to provide constructive feedback to sustain and enhance such a system.
Reliable data are essential for successful applications of data science and AI, but they are only the starting point. The next challenge is ensuring the reliability of the methods and technologies used to extract information from the data. This challenge becomes particularly daunting when methods and algorithms are developed and refined empirically, without solid theoretical underpinning or understanding. Realistic empirical studies evaluating the performance of black-box or grey-box algorithms are especially crucial in these cases, as they can provide some measure of confidence—or serve as warnings about untrustworthy algorithms.
The article by Yudi Pawitan and Chris Holmes (2025), “Confidence in the Reasoning of Large Language Models,” is therefore timely and thought-provoking. Their findings may be troubling, particularly for those of us dedicated to studying and managing uncertainties. The concern lies not only in the significant variations displayed by models such as GPT-4, GPT-4 Turbo, and Mistral, but also in their alarming overconfidence: “The LLMs [large language models] frequently report 100% confidence in their answers, even when those answers are incorrect.” The authors urge caution, advising readers to interpret LLM responses carefully, especially when the models express high confidence. They warn, “Current LLMs do not have a coherent understanding of uncertainty.” While the behavior of LLMs may never be fully understood—either by AI or humans—it is imperative that we develop sufficient insights into their extreme behaviors to establish effective guardrails. Minimally, we must constantly remind ourselves to avoid becoming AI automatons, that is, to resist uncritically trusting AI methods or systems without scrutiny.
It might surprise some readers that this issue of HDSR features five articles on wine. Data science for wine? “What are you drinking, Xiao-Li?” some may ask. Well, to paraphrase W. C. Fields, I teach with wine, sometimes I even blend it with food for thought. (No kidding, see Tiziana Alocci, 2024, and Meng, 2022.) As I mentioned in my editorial for issue 6.3 (Meng, 2024), HDSR celebrated its fifth anniversary with two symposiums: AI and Data Science: Integrating Artificial and Human Ecosystems and Vine to Mind.
After its opening session, the Vine to Mind symposium covered three broad topics:
Data-Driven Wine Economics
Data and AI for the Wine Industry
The Impact of Climate Change on the Wine Industry
These topics provide one more demonstration of why data science and AI touch on everything and are needed everywhere.
But even for those who have little interest in studying wine, there is a research rationale for choosing the wine theme to celebrate HDSR’s fifth anniversary. As I wrote in my introduction to the symposium, “From Bottle Shock to Future Shock: Everything Data Science and Data Science for Everyone” (Meng, 2025), “any data science method that can handle wine data well would likely do well in general, because of the intoxicatingly complex nature of quantitatively studying wine.” The journey from vine planting to wine consumption is long and intricate, with the ultimate consumer satisfaction influenced by myriad factors ranging from climate change to mood shifting. After all, the same wine can taste markedly different when sipped alone versus when shared in joyful company.
The advancement of AI certainly brings new varieties of methods and technologies to the wine industry and wine studies. Jing Cao’s (2025) article in the Recreations in Randomness column, “AI: A New and Impactful Player in the Quality Evaluation of Wine,” provides a timely tasting, covering topics from wine production to wine consumption. Further samples will appear in the upcoming Spring issue of HDSR, featuring selected technical articles resulting from the Vine to Mind symposium.
The Vine to Mind articles in this issue are the first of a two-part special theme, with this first part focusing on broad perspectives. While my introductory editorial (Meng, 2025) was written from a statistician’s perspective, highlighting the persuasive power of principled statistical methods through the historical 1976 Judgment of Paris, Donald St. Pierre (2025) offers “An Optimist’s View on the Future of Wine” as an industry leader. This is paired with “Wine Economics: Emergence, Topics, and Outlook” by Karl Storchmann and Jing Cao (2025), written from the vantage point of academic researchers. Despite varying perspectives, our core messages align: the enterprise of ‘vine to mind’—from wine production to wine appreciation—is a complex ecosystem. The use of data science and AI is therefore essential, but much remains to be done.
Few challenges affect ‘everything everywhere all at once’ more than climate change. Viticulture is certainly among the most impacted, as made abundantly clear in the panel discussion “Global Climate Change and Wine Production: Industry and Academic Perspectives,” featured in the Vine to Mind symposium. The panel, which included industry leaders and academic researchers Robert Stavins, Laura Catena, Elisabeth Forrestel, Jean-Baptiste Rivail, Mark Sahn, and Daniel A. Sumner (2025), addressed wide-ranging topics: water shortages, winegrowing, building resilient vines, reducing greenhouse gas emissions, blending science with tradition, assessing regional differences, and so forth.2 Contemplating the statistical challenges—especially the spatial, temporal, and cultural variabilities involved—makes my head spin. I may need to pour myself a glass of Oxalis Huatata (a birthday gift) just to pace my thoughts.
To add more perspectives on winemaking and consumption, Christian Kittery (2025) reflects on the symposium in “The Biggest Ideas for the Future of Wine: Recap of the Vine to Mind Symposium.” He begins from a cultural exchange perspective, a dimension that has received far less attention in the media and research publications than it deserves. Kittery summarizes his takeaways from the symposium in four main ideas: the revolutionary nature of AI, the essential role of psychology in marketing, the need to balance tradition and innovation, and the shared responsibility of addressing climate change.
None of the four ideas in Kittery’s (2025) summary are restricted to wine. This encapsulates the very essence of having a global forum like HDSR—to discuss challenges, remedies, and opportunities that transcend specific domains. There is much to gain by learning from others, especially when those others are diverse and plentiful. Indeed, the advances made by LLMs and similar technologies are largely due to their ability to harvest insights from an increasingly massive “wisdom of crowds,” or what I called “Homo sapiens intelligence” in my editorial for issue 5.4 (Meng, 2023). There is nothing artificial about today’s AI systems, as they are built by humans, trained by humans, and on data created, collected, curated, and controlled by humans.
As just one more example of the essential role of human perspectives, this issue also features a philosophical contemplation by Kino Zhao (2025), “What Are Statistical Assumptions About? An Answer From Perspectivism.” Making assumptions—statistical or otherwise—is an inevitable human endeavor in our quest to understand nature, ourselves, and the interactions among all of them. Frequently, we make assumptions not because they represent ground truth but because they help us organize thoughts, extract essence, and make progress or actionable plans, especially under resource constraints (e.g., data, time, funding). Zhao’s thesis formalizes these intuitive understandings from a philosophical perspective, provoking deeper reflection on how humans learn, reason, and make inferences.
Regardless of how we embrace or fear real AI, now often referred to as artificial general intelligence, perhaps we can agree that a telltale sign of artificial general intelligence would be its ability to reason from different perspectives, while appreciating their discordances and conflicts as features of reasoning rather than (algorithmic) bugs. Humans excel at identifying and creating frameworks to redirect ourselves toward constructive and productive paths. Recognizing the potential falsehood of an assumption does not necessarily prevent us from proceeding with the assumption; instead, it compels us to weigh the risks of adoption and evaluate potential trade-offs. When all factors are considered, such mental exercises may even lead to practical optimality. As I understand it, Zhao’s (2025) perspectivist framework justifies such approaches, though I wouldn’t be surprised if she needs to redirect her thoughts in order to agree that I have fairly illustrated her framework.
Whether an AI system can ever conduct such a nuanced thought experiment remains an open question. But I won’t lose sleep over it—feeling ‘everything everywhere all at once’ doesn’t leave many hours for sleep to lose. Jing Cao (2025) concludes her column about how AI is reshaping wine evaluation with the sentiment that “Surely, AI can render great help in some aspects, but some areas of experience might be best left to ourselves.” She was referring to making, buying, and tasting wine. With a birthday wine-tasting event eagerly awaited, I find no need to redirect my thoughts in order to share Cao’s preference. Instead, I raise a toast to Zhao’s perspectivism—let’s keep that to humans.
Xiao-Li Meng has no financial or nonfinancial disclosures to share for this editorial.
Auerbach, J., McKay Bowen, C., Citro, C. F., Pierson, S., & Seeskin, Z. H. (2025). The nation’s data at risk: The first annual report on the federal statistical system. Harvard Data Science Review, 7(1). https://doi.org/10.1162/99608f92.ca01ab52
Cao, J. (2025). AI: A new and impactful player in the quality evaluation of wine. Harvard Data Science Review, 7(1). https://doi.org/10.1162/99608f92.be44bdad
Duijm, P., & van Lelyveld, I. (2025). Data Science for central banks and supervisors: How to make it work, actually. Harvard Data Science Review, 7(1). https://doi.org/10.1162/99608f92.4d41c5a4
Guo, Z., Ding, C., Do, D., Shah, A., Lee, R. J., Hu, X., & Rudin, C. (2025). Improving atrial fibrillation detection using a shared latent space for ECG and PPG signals. Harvard Data Science Review, 7(1). https://doi.org/10.1162/99608f92.9e63a630
Hamutcu, H. (2025). Introduction to special theme on data literacy at scale. Harvard Data Science Review, 7(1). https://doi.org/10.1162/99608f92.09ad573b
James, T., & Hamutcu, H. (2025). Navigating the rapid evolution of AI with Todd James. Harvard Data Science Review, 7(1). https://doi.org/10.1162/99608f92.90142312
Hilty, E. C., Vilski, V., Mishra, S., Condon, P., & Gracia, S. M. (2025). Building an inclusive data literacy community. Harvard Data Science Review, 7(1). https://doi.org/10.1162/99608f92.d622eaff
Kittery, C. (2025). The biggest ideas for the future of wine: Recap of the Vine to Mind symposium. Harvard Data Science Review, 7(1). https://doi.org/10.1162/99608f92.84acc1c1
Koloski, D., Porter, C., Almand-Hunter, B., Gatchell, S., & Logan, V. (2025). Data literacy in industry: High time to focus on operationalization through middle managers. Harvard Data Science Review, 7(1). https://doi.org/10.1162/99608f92.6f5dfc6f
Kwan, D., & Scheinert, D. (Directors). (2022). Everything everywhere all at once [Film]. A24; IAC Films; Gozie Agbo; Year of the Rat; Ley Line Entertainment.
Meng, X.-L. (2022). Seeking simplicity in statistics, in wine, and everything else in fortune cookies. FONDATA, (3), 19–27. https://www.dropbox.com/scl/fi/3l1w9qqlllnus8uuizt6p/Wine-and-Statistics-V4.pdf?rlkey=ndy38dbzds3grlgarww9cdccp&e=1&dl=0
Meng, X.-L. (2023). Human intelligence, artificial intelligence, and Homo sapiens intelligence? Harvard Data Science Review, 5(4). https://doi.org/10.1162/99608f92.11d9241f
Meng, X.-L. (2024). From COVID-19 to GPT-4o: The groundbreaking quinquennium for Harvard Data Science Review (and humanity). Harvard Data Science Review, 6(3). https://doi.org/10.1162/99608f92.f3abeae2
Meng, X.-L. (2025). From bottle shock to future shock: Everything data science and data science for everyone. Harvard Data Science Review, 7(1). https://doi.org/10.1162/99608f92.aab45e43
Pawitan, Y., & Holmes, C. (2025). Confidence in the reasoning of large language models. Harvard Data Science Review, 7(1). https://doi.org/10.1162/99608f92.b033a087
Srivastava, A., & Hamutcu, H. (2025). In conversation with Intuit Chief Data Officer Ashok Srivastava. Harvard Data Science Review, 7(1). https://doi.org/10.1162/99608f92.f250e959
St. Pierre, D. (2025). An optimist’s view on the future of wine. Harvard Data Science Review, 7(1). https://doi.org/10.1162/99608f92.56d4c11d
Stavins, R. N., Catena, L., Forrestel, E., Rivail, J.-B., Sahn, M., & Sumner, D. A. (2025). Global climate change and wine production: Industry and academic perspectives. Harvard Data Science Review, 7(1). https://doi.org/10.1162/99608f92.a17caeae
Storchmann, K., & Cao, J. (2025). Wine economics: Emergence, topics, and outlook. Harvard Data Science Review, 7(1). https://doi.org/10.1162/99608f92.c94174e8
Tiziana Alocci. (2024, March 21). Fondata Magazine – Visualising wine tasting data. https://www.tizianaalocci.com/home/visualizing-wine-data
Zhao, K. (2025). What are statistical assumptions about? An answer from perspectivism. Harvard Data Science Review, 7(1). https://doi.org/10.1162/99608f92.7fcd521d
©2025 Xiao-Li Meng. This editorial is licensed under a Creative Commons Attribution (CC BY 4.0) International license, except where otherwise indicated with respect to particular material included in the editorial.