This Software Worked Fine Earlier – What Happened? Architectural Archipelagos Are Impediments to Scientific Progress

Speaker:  Nenad Medvidović – Los Angeles, USA
Topic(s):  Human Computer Interaction , Software Engineering and Programming , Information Systems, Search, Information Retrieval, Database Systems, Data Mining, Data Science

Abstract

Modern software has become indispensable but continues to present a bewildering array of unexpected, and often undesirable, behaviors. All of us have been induced to update our operating system, only to find that a favorite application no longer works the same (or, in extreme instances, at all). This is even more pronounced in research settings, where the software produced by related projects has tended to span “archipelagos” of tools with often unpredictable properties. Archipelagos emerge through a recurring multi-year evolution chain of the tools’ underlying architectures: one researcher builds a helpful utility, another subsequently “tweaks” it to fit their need, a third uses a portion of the original tool and adds an unrelated capability for their project, and so on. Such architectural archipelagos are especially prevalent in academic settings, where proofs-of-concept are frequently developed to run experiments that tend not to generalize and then discarded after they fulfill their purpose (e.g., when a submitted paper is accepted or a Ph.D. dissertation completed), or they evolve into loose collections of components, tools, frameworks, workbenches, and/or environments (e.g., when the existing capabilities are viewed as a good foundation for a new research project). These archipelagos possess unique characteristics and suffer from unique complications: due to the haphazard processes by which they emerge, they inherently accumulate technical debt. In turn, this directly hampers their transition to other research groups or to industrial usage, despite containing state-of-the-art technology. The duplication of effort archipelagos engender slows the pace of scientific progress, particularly exacerbated in light of the current explosion of AI-related research. This talk will explain the archipelago model, how and why it tends to emerge especially in academic-research settings, illustrate it with real-world examples, discuss the lessons-learned in the process of both growing and trying to make use of others’ archipelagos, and suggest some paths forward. 

About this Lecture

Number of Slides:  50
Duration:  45 - 50 minutes
Languages Available:  English
Last Updated:  23/03/2026

Request this Lecture

To request this particular lecture, please complete this online form.

Request a Tour

To request a tour with this speaker, please complete this online form.

All requests will be sent to ACM headquarters for review.