What you can read below, is the text I wrote after ICSE 2013. Unfortunately, three years later the situation is exactly the same. And it’s not just me who says this.
From the “Insights and Lessons Learned from Analyzing ICSE 2016 Survey and Review Data” report written the ICSE 2016 PC Chairs
There were 3 topics that had no papers accepted: Autonomic and (self) adaptive systems; Model-driven engineering and Specification; and modeling languages. Note that each of these topics had a fair number of submissions.
and
What might need some attention is the fact that Model-driven engineering for the 2nd year in a row got zero papers accepted (out of 40 submitted the last 2 years).
While the ICSE community is having deep discussions on the announced ICSE submission cap, I think a discussion on why some topics are under represented at ICSE also deserves the attention of the community. Otherwise, there is the risk ICSE ends up being ICS with a close focus on program analysis, testing and software verification topics.
(and before you start scanning the titles of this year and suggest that there were indeed a few papers on modeling, two things: 1 – we are talking about technical research papers and 2 – any technical research paper you see you believe it could be a modeling paper was not classified as such by the paper authors).
Enter myself three years ago.
The program for the ICSE 2013 conference is now online .
To be honest, I’m very disappointed by the list of accepted papers. Once again, papers on modeling (our sense of the word modeling, i.e. model checking papers do not count) are clearly the exception. I don’t really understand why; it’s fine that not all software engineers believe in the full and automatic generation of software from models (i.e. executable models) but model-driven engineering is much more than this and covers the use of models in other software engineering tasks (like reverse engineering) with different degrees of formalism and precision. I can’t believe that any software engineer would argue than creating a model of a system is a bad practice.
As such, if we all agree that modeling (in the broadest sense of the word) is an important element in software engineering, why is not better represented in the (supposedly) best research conference in software engineering? I have no idea. Surprisingly, the MiSE workshop (Modeling in Software Engineering), co-located with ICSE has been around for several years now and seems to be quite successful (I’ve attended a couple of editions but of course my opinion is not objective) but this does not results in more modeling papers in the main conference.
On a side note, my feeling is that the papers are more about the “software” than the “engineering” side of software engineering. Only with the titles is difficult to provide a precise analysis but I don’t see many papers dealing with the engineering aspects of software development/maintenance…
Of course, maybe I’m just angry because my two submissions were rejected. We’ll try again next year, hoping to fill one of the few “modeling paper” slots that will be available (based on the observed trend).
FNR Pearl Chair. Head of the Software Engineering RDI Unit at LIST. Affiliate Professor at University of Luxembourg. More about me.
(comment by Jon Whittle)
Jordi: there are lots of reasons for this. As someone who served on the ICSE PC, I’d be happy to chat about it sometime. Off the top of my head though, there are a few things that could be the reason. First, the MODELS community isn’t submitting to ICSE. Or at least not in the papers discussed in the PC meeting. Yes, there were some, but they were dwarfed by papers on testing. Which brings me to: a large number of papers discussed were on testing. Perhaps because it is easier to provide a rigorous evaluation — data is available and comparison with prior techniques is possible. I think we need a much more concerted effort if we want the MODELS community to send papers to ICSE — they need to understand better what is required from an ICSE paper, but we also need, as a community, to have better ways of objectively evaluating our work.
Jon, I agree with the evaluation part. The current trend of emphasizing the evaluation part in SE papers is a problem for “us” since we are not so used to give a lot of importance to this
Jon’s comment/observation sounds right to me. At MoDELS very often you can “get away” with evaluation via a case study, which — if done right — is appropriate and useful, but sometimes is just there to give proof of concept. ICSE needs something much more mature.
I think there’s also a cost-benefit analysis being applied here. The amount of effort required to “ICSE”fy a paper is significant. I often prefer to “journal”fy the paper instead (the amount of effort is not too dissimilar) and save on the travel expense as well as benefit from the conversation you can have with journal reviewers/editors that you can’t easily have with conference reviewers.
Jordi, I have been checking the list of papers and also got a little bit surprised … I found it weird not to see a dedicated “track/session” for Modeling (?). As a side note, during last years there was always a “Product-lines” session and a bunch of papers on product-lines. It seems to me that modeling shall be at least similarly represented. Reasons mentioned in the comments by Jon/Richard sound right to me too.
I scanned the titles (and a few of the abstracts) to find these topics, which all seem related to MDE:
Managing Non-functional Uncertainty via Model-Driven Adaptivity
Learning Revised Models for Planning in Adaptive Systems
Data Clone Detection and Visualization in Spreadsheets
Aluminum: Principled Scenario Exploration through Minimality
Counter Play-Out: Executing Unrealizable Scenario-Based Specifications
Unifying FSM-Inference Algorithms through Declarative Specification
Dynamic Injection of Sketching Features into GEF Based Diagram Editors
Discovering Essential Code Elements in Informal Documentation
UML in Practice
The Role of Domain Knowledge and Hierarchical Control Structures in Socio-technical Coordination
Requirements Modelling by Synthesis of Deontic Input-Output Automata
Perhaps I cast the net a little wide, but you suggested that modeling is a wide topic, and I agree.
Will
Hello,
I agree with all the above. My feeling (and it is just a personal feeling because I have only anecdotal evidence to support that) is that we are in paradoxal situation. On the one hand, there is no “modelling” paper tagged as such in the technical program as reported by the ICSE PC chairs and Jordi, and on the other hand, Will reported several that actually use models for software engineering activities. The product line session (in which I co-authored a paper) had two papers exclusively working with models (but not tagged as such): Featured Model-based Mutation Analysis (ours) and Feature-Model Interfaces: The highway to Compositional Analyses of Highly-Configurable Systems. We had to clarify in the rebuttal why we were doing model-based mutation (this is still a relatively new area for mutation testing, which has much more experience with code-based approaches) so I don’t know whether whether it was a bias against modelling approaches or a comment on the respective popularity of approaches.
I also attended the V2025 session where Lionel Briand and Pamela Zave presented cases advocated the use of stochastic models to test cyberphysical systems or the need for modelling domains (which can been seen as promoting DSMLs). This last vision was awarded.
So we can see the glass half-empty as you do. But I have the will to see it also half-full: modelling approaches have pervaded all tracks of ICSE, and are part of the visionary discussions on software engineering.
Interestingly, there is a parallel with software product lines: a few years ago, there was only one venue for such papers, namely the software product line conference. Now we see them in RE, ICSE, FSE or MODELS not necessarily with “software product line” keyword, sometimes replaced by variability-intensive, highly-configurable ones. There are normally differences amongst these keywords but in practice the papers do not always acknowledge that and most of the analyses developed apply to any of these keywords. I see this as a mark of popularity for these approaches, though there is not always a product line session in all these venues.
Based on that, I see hope for MDE at ICSE even if this means hard times for bibliometrics 🙂
Gilles.
Always nice to have an optimistic in the house! Though I wonder why authors did not classify their papers as modeling papers (maybe they were scared to do so?). Regarding Visio 2025, this is a separate track. I agree in these other tracks (like SEIS or SEIP) we can find models papers but we don’t have the same proportion of them in the pure technical track. I do find this strange
I participated in workshops and main tracks. I was so disappointed because instead of seeing people working on engineering parts of software (or papers presented on MDE) I ended up listening to results of work done by data scientists who just applied ML algorithms to data. As a software engineer I believe we need those results to improve methods and processes, but it doesn’t mean we are done with development methods like MDE that can bring more benefits to the community. I noticed ICSE PCs got affected with the general popularity of ML and think any work with ML is new and valuable. However, they forgot it’s not software engineering.
Yes, some of you indicated about the evaluation problems in MDE but all of the papers published based on data also have the evaluation problem. Why? Because all of those results can be completely wrong in my own data, team, company, etc. Therefore, how they accept those papers but don’t accept our papers which come with case studies and formal definitions.
P.s. I think when top people in SE get fund from companies to come and alnayze their data so they have to accept these kinds of papers. If you check paper authors you may end up with this point of view as well.
Even though I was selected to present my MDD Environment within a demo session, I agree… MDE should have more attention within a conference with such a broad range of topics.