AI coding agents contribute thousands of pull requests daily to open-source repositories, yet practitioners lack empirical guidance on optimal contribution sizing. We adapt the therapeutic index framework from pharmacology to characterise the relationship between pull request size and integration outcomes. Analysing 33,078 agent-authored pull requests from the AIDev dataset across five AI coding agents, we model dose-response relationships for efficacy (merge probability) and toxicity (review friction). All agents exhibit statistically significant negative relationships between size and merge probability (p < 0.001), differing substantially in baseline performance (47.5% to 82.6%) and dose sensitivity. Task type substantially moderates these relationships: bug fixes exhibit ED50 of 1,467 lines [95% CI: 1,025–2,216] with therapeutic index of 35.2, whilst features show ED50 of 18,234 lines [95% CI: 12,337–28,390] with therapeutic index of 2.3. This 15-fold difference, confirmed by non-overlapping confidence intervals, indicates that bug fixes exhibit wider therapeutic windows whilst feature implementations exhibit narrower windows.

Context: Open-source software (OSS) development is often studied as a decentralized process driven by technical goals. However, mature OSS projects operate under external constraints such as security advisories, release deadlines, and ecosystem dependencies. These pressures shape technical decisions and also communication patterns among contributors, including emotional expression. Objective: This study investigates how emotional expression in OSS projects varies across different types of repositories, evolves over time, and relates to the activity of top contributors. The goal is to assess whether emotional dynamics are shaped more by project function than by technical domain or project size. Methods: We analyzed issue comments from 14 OSS repositories spanning over ten years. A transformer-based classifier was used to detect emotions. Emotional patterns were quantified using a composite Emotional Index, and contextual activity. Contributor roles were assessed using a Contribution Index combining code activity, discussion engagement, and sustained involvement. Analyses were conducted at the repository, temporal, and contributor levels. Results: The four most frequent emotions across all repositories were gratitude, curiosity, confusion, and approval. Emotional patterns tend to cluster by functional role rather than technical domain, with repositories converging toward stable emotional profiles over time. High-impact contributors show distinct expression patterns that reflect their role and stage of engagement. Conclusion: Emotional expression in OSS projects follows recurring patterns linked to project function, contributor roles, and maturity. These findings can help anticipate communication challenges during project evolution and support interaction strategies among contributor groups with differing emotional tendencies.

Developers inevitably make human errors while coding. These errors can lead to faults in code, some of which may result in system failures. It is important to reduce the faults inserted by developers as well as fix any that slip through. To investigate the fault insertion and fault fixing activities of developers. We identify developers who insert and fix faults, ask whether code topic ‘experts’ insert fewer faults, and experts fix more faults and whether patterns of insertion and fixing change over time. We perform a time-based analysis of developer activity on twelve Apache projects using Latent Dirichlet Allocation (LDA), Network Analysis and Topic Modelling. We also build three models (using Petri-net, Markov Chain and Hawkes Processes) which describe and simulate developers’ bug-introduction and fixing behaviour. We show that: the majority of the projects we analysed have developers who dominate in the insertion and fixing of faults; Faults are less likely to be inserted by developers with code topic expertise; Different projects have different patterns of fault inserting and fixing over time. We recommend that projects identify the code topic expertise of developers and use expertise information to inform the assignment of project work.

Smart Contracts are general-purpose programs that provide a higher level of security than traditional contracts and reduce other transaction costs associated with the bargaining practice, as they are executed in a Blockchain infrastructure. Developers use smart contracts to build their tokens and set up gambling games, crowd sales, ICO, and many others domains of application. The security of Smart Contracts is also crucial, as SCs at the very core level, move money. In recent years, researchers have provided a set of known vulnerabilities that afflict SCs. This study analyzed the relationship between the SC domain of application, domain category, and known vulnerabilities. We categorized the SC using the topic modeling on a curated dataset of SC annotated with know vulnerabilities. Indeed, we found that a certain category of SC is strongly associated with specific vulnerabilities.

This work shows the results obtained from a comparison between a restricted and a unrestricted Bitcoin price classification, verifying whether the addition of technical indicators to the classic macroeconomic variables leads to an effective improvement in the prediction of Bitcoin price changes. The goal was achieved implementing different machine learning algorithms, such as Support Vector Machine (SVM), XGBoost (XGB), a Convolutional Neural Network (CNN) and a Long Short Term Memory (LSTM) neural network. Macroeconomic variables data were gained from Yahoo Finance website spanning a 4-year interval with a hourly resolution, while technical indicators data are provided by the python talib library. The variance problem on test samples has been taken into account through the cross validation technique which also allowed to evaluate a more reliable estimate of the model's performance. Furthermore, the Grid Search technique was used to find the best hyperparameters values for each implemented algorithm. The results were evaluated in terms of the well known classification metrics, i.e. accuracy, precision, recall and f1 score. Based on the results, it was possible to demonstrate that the unrestricted case outperforms the restricted one, verifying that the addition of the technical indicators to the macroeconomic variables actually improves the accuracy on Bitcoin price classification.

Similar to other industries, the software engineering domain is plagued by psychological diseases such as burnout, which lead developers to lose interest, exhibit lower activity and/or feel powerless. Prevention is essential for such diseases, which in turn requires early identification of symptoms. The emotional dimensions of Valence, Arousal and Dominance (VAD) are able to derive a person's interest (attraction), level of activation and perceived level of control for a particular situation from textual communication, such as emails. As an initial step towards identifying symptoms of productivity loss in software engineering, this paper explores the VAD metrics and their properties on 700,000 Jira issue reports containing over 2,000,000 comments, since issue reports keep track of a developer's progress on addressing bugs or new features. Using a general-purpose lexicon of 14,000 English words with known VAD scores, our results show that issue reports of different type (e.g., Feature Request vs. Bug) have a fair variation of Valence, while increase in issue priority (e.g., from Minor to Critical) typically increases Arousal. Furthermore, we show that as an issue's resolution time increases, so does the arousal of the individual the issue is assigned to. Finally, the resolution of an issue increases valence, especially for the issue Reporter and for quickly addressed issues. The existence of such relations between VAD and issue report activities shows promise that text mining in the future could offer an alternative way for work health assessment surveys.

Issue tracking systems store valuable data for testing hypotheses concerning maintenance, building statistical prediction models and (recently) investigating developer affectiveness. For the latter, issue tracking systems can be mined to explore developers emotions, sentiments and politeness---affects for short. However, research on affect detection in software artefacts is still in its early stage due to the lack of manually validated data and tools.

In this paper, we present an analysis of more than 500 K comments from open-source repositories of software systems developed using agile methodologies. Our aim is to empirically determine how developers interact with each other under certain psychological conditions generated by politeness, sentiment and emotion expressed within developers’ comments. Developers involved in an open-source projects do not usually know each other; they mainly communicate through mailing lists, chat, and tools such as issue tracking systems. The way in which they communicate affects the development process and the productivity of the people involved in the project. We evaluated politeness, sentiment and emotions of comments posted by agile developers and studied the communication flow to understand how they interacted in the presence of impolite and negative comments (and vice versa ). Our analysis shows that “firefighters” prevail. When in presence of impolite or negative comments, the probability of the next comment being impolite or negative is 13 % and 25 %, respectively; ANGER however, has a probability of 40 % of being followed by a further ANGER comment. The result could help managers take control the development phases of a system, since social aspects can seriously affect a developer’s productivity. In a distributed agile environment this may have a particular resonance.

Human Affectiveness, i.e., The emotional state of a person, plays a crucial role in many domains where it can make or break a team's ability to produce successful products. Software development is a collaborative activity as well, yet there is little information on how affectiveness impacts software productivity. As a first measure of this impact, this paper analyzes the relation between sentiment, emotions and politeness of developers in more than 560K Jira comments with the time to fix a Jira issue. We found that the happier developers are (expressing emotions such as JOY and LOVE in their comments), the shorter the issue fixing time is likely to be. In contrast, negative emotions such as SADNESS, are linked with longer issue fixing time. Politeness plays a more complex role and we empirically analyze its impact on developers' productivity.

Software development is a collaborative activity in which developers interact to create and maintain a complex software system. Human collaboration inevitably evokes emotions like joy or sadness, which can affect the collaboration either positively or negatively, yet not much is known about the individual emotions and their role for software development stakeholders. In this study, we analyze whether development artifacts like issue reports carry any emotional information about software development. This is a first step towards verifying the feasibility of an automatic tool for emotion mining in software development artifacts: if humans cannot determine any emotion from a software artifact, neither can a tool. Analysis of the Apache Software Foundation issue tracking system shows that developers do express emotions (in particular gratitude, joy and sadness). However, the more context is provided about an issue report, the more human raters start to doubt and nuance their interpretation of emotions. More investigation is needed before building a fully automatic emotion mining tool.