The Essence of Software Engineering

Every day, we see more examples in the media of discrimination in software. Evidence of software discrimination has been found in modern software systems that recommend criminal sentences, grant access to financial products, and determine who is allowed to participate in promotions. This talk defines software fairness and discrimination and develops a testing-based method for measuring if and how much software discriminates, focusing on causality in discriminatory behavior. Our approach, Themis, automatically generates efficient test suites to measure discrimination. A key observation is that measuring discrimination requires causal reasoning. Themis uses a new kind of testing -- causal testing -- to measure causal relationships between sensitive inputs and software behavior. That is, Themis can tell if a specific sensitive input is responsible for a particular software behavior. This enables Themis to test for types of bias that better capture our intuitive notion of undesirable discrimination in software. Evaluating Themis on open-source machine learning systems finds that Themis is effective at discovering software discrimination, state-of-the-art techniques for removing discrimination from algorithms fail in many situations, and Themis optimizations are effective at producing efficient test suites for measuring discrimination. The talk will argue that discrimination testing is a critical aspect of the software development cycle in domains with possible discrimination and describe initial tools for measuring software discrimination.

Dr. Ivar Jacobson will revisit the history of methods with method wars and fashionable method frameworks. He will explain why we need to break out of our repetitive dysfunctional behaviour, and introduce Essence: a new way of thinking that promises to free the practices from the methods and enable true learning organisations. Free the practices so that teams can experiment, innovate and plug and play with proven practices to create their own way-of-working that they need today and seamlessly evolve into the one they need tomorrow.

Blockchains introduce a set of key characteristics revolving around decentralized control and trustless interactions. These make them transformative for the next generation of business applications. A “blockchain-aware” software architecture must not compromise but fully align with these characteristics. Ensuring this, however, is far from being trivial. This talk explores important architectural needs when engineering blockchain-based systems – focusing on off-chaining data and computation – and discusses ways to approach them.

Software process models are in a shambles. At one end, CMMI and such help assess adherence to codified management rules but impose bureaucratic strangleholds on projects and programmers. At the other, agile methods rehabilitate the importance of code but do little to promote adoption of modern approaches to software quality. Variants abound, such as the spiral model, but they are subject to endless interpretations. Sooner or later model fatigue sets in and teams start doing whatever suits their needs, filling the forms, if necessary, to placate authorities but not gaining much from the theoretical models. This situation is unhealthy since a serious engineering approach needs to rely on a widely accepted process model that fosters that is precise enough to help teams achieve defined goals and flexible enough to let each project adapt to its specific circumstances. It is time for a fresh approach. The new ‘All-In-One’ process model presented in this talk, starting from the author’s earlier work on seamless development, the Eiffel method and language, the Single Product Principle and the Cluster lifecycle model, includes a comprehensive set of techniques from the forefront of software engineering advances into a single framework.

For large distributed partly embedded software systems, a design concept is presented supporting composition, parallelism, and concurrency and finally real time but keeping all of advantages of object-oriented programming. The approach to model, specify and implement systems provides an execution model which is parallel and concurrent in nature and supports real time and composition leading to a system engineering style structuring software systems in terms of functional features, quality attributes, and contracts.

Change makes software different from any other artifact. Although this is known since the 1970s, change is still often handled in an ad-hoc manner. Agile development and, more recently, DevOps have been proposed as a solution, and success stories are reported from industry. Still, principled and rigorous foundations that can be taught, practiced, and systematically replicated are lacking. We distinguish between evolution and adaptation and discuss how software can be made self-adaptive, and how it can cooperate with humans ‘in-the-loop’ to support change. We outline a roadmap of future work needed to progress in tool support and automation to lead to dependable adaptation and evolution.