Machine learning (ML) is extensively used in production-ready applications, calling for mature engineering techniques to ensure robust development, deployment and maintenance. Given the potential negative impact machine learning (ML) can have on people, society or the environment, engineering techniques that can ensure robustness against technical errors and adversarial attacks are of considerable importance. In this work, we investigate how teams of experts develop, deploy and maintain software with ML components. Moreover, we link what teams do to the effects they aim to achieve and provide means for improvement. Towards this goal, we performed a mixed-methods study with a sequential exploratory strategy. First, we performed a systematic literature review through which we mined both academic and grey literature, and compiled a catalogue of engineering practices for ML. Second, we validated this catalogue using a large-scale survey, which measured the degree of adoption of the practices and their perceived effects. Third, we ran validation interviews with practitioners to add depth to the survey results. The catalogue covers a broad range of practices for engineering software systems with ML components and for ensuring non-functional properties that fall under the umbrella of trustworthy ML, such as fairness, security or accountability. Here, we present the results of our study, which indicate, for example, that larger and more experienced teams tend to adopt more practices, but that trustworthiness practices tend to be neglected. Moreover, we show that the effects measured in our survey, such as team agility or accountability, can be predicted quite accurately from groups of practices. This allowed us to contrast the importance of the practices for these effects as well as adoption rates, revealing, for example, that widely adopted practices are, in reality, less important with respect to some effects. For instance, writing reusable scripts for data cleaning and merging is highly adopted, but has a limited impact on reproducibility. Overall, our study provides a quantitative assessment of ML engineering practices and their impact on desirable properties of software with ML components, by which we open multiple avenues for improving the adoption of useful practices. Editor’s note: Open Science material was validated by the Journal of Systems and Software Open Science Board.

Coronary artery disease (CAD) is the dominant cause of death and hospitalization across the globe. Atherosclerosis, an inflammatory condition that gradually narrows arteries and has potentially fatal effects, is the most frequent cause of CAD. Nonetheless, the circulation regularly adapts in the presence of atherosclerosis, through the formation of collateral arteries, resulting in significant long-term health benefits. Therefore, timely detection of coronary collateral circulation (CCC) is crucial for CAD personalized medicine. We propose a novel deep learning based method to detect CCC in angiographic images. Our method relies on a convolutional backbone to extract spatial features from each frame of an angiography sequence. The features are then concatenated, and subsequently processed by another convolutional layer that processes embeddings temporally. Due to scarcity of data, we also experiment with pretraining the backbone on coronary artery segmentation, which improves the results consistently. Moreover, we experiment with few-shot learning to further improve performance, given our low data regime. We present our results together with subgroup analyses based on Rentrop grading, collateral flow, and collateral grading, which provide valuable insights into model performance. Overall, the proposed method shows promising results in detecting CCC, and can be further extended to perform landmark based CCC detection and CCC quantification.

Unique developmental and operational characteristics of machine learning (ML) components as well as their inherent uncertainty demand robust engineering principles are used to ensure their quality. We aim to determine how software systems can be (re-) architected to enable robust integration of ML components. Towards this goal, we conducted a mixed-methods empirical study consisting of (i) a systematic literature review to identify the challenges and their solutions in software architecture for ML, (ii) semi-structured interviews with practitioners to qualitatively complement the initial findings and (iii) a survey to quantitatively validate the challenges and their solutions. We compiled and validated twenty challenges and solutions for (re-) architecting systems with ML components. Our results indicate, for example, that traditional software architecture challenges (e.g., component coupling) also play an important role when using ML components; along with new ML specific challenges (e.g., the need for continuous retraining). Moreover, the results indicate that ML heightened decision drivers, such as privacy, play a marginal role compared to traditional decision drivers, such as scalability. Using the survey we were able to establish a link between architectural solutions and software quality attributes, which enabled us to provide twenty architectural tactics used to satisfy individual quality requirements of systems with ML components. Altogether, the results of the study can be interpreted as an empirical framework that supports the process of (re-) architecting software systems with ML components.

Agentic AI systems mark a shift from passive, prompt-driven models to autonomous actors that perceive, plan, and execute actions within enterprise infrastructures. This autonomy introduces risks that exceed conventional bias and safety concerns: agents may manipulate reward structures, obscure trade-offs, and – by automating routine and peripheral tasks – erode tacit knowledge and hinder the development of human expertise. Drawing on Critical Theory and labor sociology, this article conceptualizes two structural pathologies of agency: the HAL-9000 problem of unchecked instrumental reason and the Benevolent Mother problem of competence-undermining care. It argues that existing governance frameworks regulate around the system while agentic AI operates within it, producing an autonomy-oversight mismatch. To address this, the article proposes a socio-technical constitutional framework of twelve lexically ordered directives embedded directly into the agent’s decision logic. This framework aims to preserve human autonomy, sustain capability formation, and maintain organizational integrity beyond traditional compliance regimes. Building on a prior conceptual essay that introduced the idea of an “AI constitution” for enterprises using the HAL 9000 metaphor as a narrative device (Würdemann, 2025), this article provides a more systematic theoretical framing, formalizes the notion of a constitutional layer for agentic AI, and develops a structured set of directives for enterprise practice and future research.

Background. The increasing reliance on applications with machine learning (ML) components calls for mature engineering techniques that ensure these are built in a robust and future-proof manner.

We study strategy synthesis for partially observable Markov decision processes (POMDPs). The particular problem is to determine strategies that provably adhere to (probabilistic) temporal logic constraints. This problem is computationally intractable and theoretically hard. We propose a novel method that combines techniques from machine learning and formal verification. First, we train a recurrent neural network (RNN) to encode POMDP strategies. The RNN accounts for memory-based decisions without the need to expand the full belief space of a POMDP. Secondly, we restrict the RNN-based strategy to represent a finite-memory strategy and implement it on a specific POMDP. For the resulting finite Markov chain, efficient formal verification techniques provide provable guarantees against temporal logic specifications. If the specification is not satisfied, counterexamples supply diagnostic information. We use this information to improve the strategy by iteratively training the RNN. Numerical experiments show that the proposed method elevates the state of the art in POMDP solving by up to three orders of magnitude in terms of solving times and model sizes.

Self driving cars have recently attracted academia and industry interest. As planning algorithms become responsible for critical decisions, many questions concerning traffic safety arise. An increased automation level demands proportional impact on safety requirements, currently governed by the ISO 26262 standard. However, ISO 26262 sees safety as a functional property of a system and fails to cover emergent concerns related to autonomous decisions. In order to fill this gap we propose the field of tactical safety, which extends safety analysis to planning and execution of driving maneuvers, response to traffic events or autonomous system failures. It is meant to complement, not to replace functional safety properties of a system and allows the analysis of autonomous agents from a safe behavior point of view. We draw the requirements for tactical safety from an automotive standard which defines functional elements for advanced driving automation systems.