Feedback Loops: The Common Thread in Agile, AI, and Life

Introduction

When I was undertaking studies in education, the concept of action research was explained. Very briefly, action research is a process for continuously improving one’s teaching methods and effectiveness over time (Parsons & Brown, 2002). It dawned on me how action research had similar processes and goals as the Agile development methodology (Martin, 2003) particularly with regard to iterations, retrospectives, and continuous improvement.

So, this led me to looking more closely at Action Research and the Agile software development methodology to identify these similarities. It turns out they share several core principles and processes that make them similar in many ways.?

Iterative Cycles

• Action Research involves a cyclical process of planning, acting, observing, and reflecting. Teachers implement a change, collect data on its effects, analyse the results, and then plan the next step based on what they learned.
• Agile follows an iterative cycle where work is divided into sprints or iterations. Each sprint involves planning, executing, reviewing, and adapting based on feedback and results.

Feedback and Adaptation

• Action Research heavily relies on feedback from students. The feedback gathered during the observation phase is necessary for making informed decisions about future actions.
• Agile emphasises regular feedback from stakeholders and team members. After each sprint, a review or retrospective meeting is held to discuss what went well, what didn’t, and what can be improved for the next iteration.

Continuous Improvement

• Action Research aims for ongoing improvement in teaching practices or educational interventions. The reflection phase allows educators to continually refine their methods and strategies.
• Agile focuses on continuous improvement in product development and team processes. Each iteration aims to deliver better value and improve the efficiency and effectiveness of the team.?

In practical terms, action research in education might involve a teacher implementing a new instructional strategy in the classroom, collecting student feedback and performance data, analysing the results, and then adjusting the strategy accordingly in a continuous loop (McNiff, 2013). Similarly, in agile software development, a development team might release a new feature in a software application, gather user feedback, analyse the usage data, and then refine the feature or develop new ones based on that feedback in successive sprints (Cockburn, 2002).?

Background

Feedback Loops are Everywhere!

It became apparent that many processes follow a similar pattern, and one of the key features of these processes is the feedback loop. I can’t remember where I first heard the term “feedback loop”, it was a long time ago, but I recall it was in reference to electrical circuits.

Electronic feedback loops are used to control the output of electronic devices, such as amplifiers. A feedback loop is created when all or some portion of the output is fed back to the input (Horowitz, 1980).

Not surprisingly, on further investigation feedback loops turn out to be fundamental components of many systems, both natural and artificial (Breewood, 2018). They play a critical role in maintaining stability, driving growth, and facilitating continuous improvement. From biological processes that regulate bodily functions (El-Samad, 2021) to technological systems that adapt and learn (McGregor, 2022) feedback loops are integral to ensuring systems operate efficiently and effectively. In fact, they are everywhere!

Positive and Negative Feedback Loops

Before going further, consider the two main types of feedback loops: positive and negative.

Positive Feedback Loops:

Positive feedback loops amplify changes, until an endpoint is reached. However, there is potential for instability. Without checks and balances, positive feedback loops can lead to runaway effects, potentially destabilizing the system.

Negative Feedback Loops:

Negative feedback loops maintain stability by counteracting changes. For instance, a thermostat regulating room temperature ensures it remains within a specific range. These loops are essential for systems requiring consistent performance and homeostasis (Wakim, 2020). In biological systems, negative feedback loops help regulate body temperature, blood sugar levels, and other vital functions to maintain equilibrium.

Discussion

Feedback Loops and Their Role in Agile Methodologies

It’s not difficult to see how feedback loops apply to agile software development. Iterations (or sprints) are designed as feedback loops to continuously improve the software product based on stakeholder input and performance metrics (Smith, 2015). At the beginning of an iteration, the team defines what needs to be achieved based on user stories and backlog items. During the iteration, the team captures feedback from ongoing development and testing. At the end of the iteration, a sprint review meeting is held to analyse the results and gather feedback from stakeholders. Insights from the analysis are used to adjust the backlog and plan the next iteration, ensuring continuous improvement.

Feedback Loops in AI and Machine Learning

Feedback loops are used in artificial intelligence (AI) and machine learning for the continuous improvement of models (Adams, 2023). During training, models use back-propagation, a feedback mechanism where errors are fed back into the model to adjust weights and biases, improving accuracy through a positive feedback loop. Additionally, systems identify and correct inconsistencies in predictions using a negative feedback loop, stabilizing the model’s performance over time. AI systems also utilise feedback from their environment to learn and adapt; for example, a recommendation system continuously improves its suggestions by using user interactions as feedback.

The End is Nigh?

In the movie "The Terminator" (Frakes, 1991) the creation of SkyNet by Cyberdyne Systems might illustrate an unchecked positive feedback loop. SkyNet, an artificial intelligence (AI) designed for defence purposes, rapidly grew beyond human control due to a lack of checks and balances, leading to the creation of Terminators and a global catastrophe. I do not want to be one of those people proclaiming “the end is nigh” as AI emerges—there are enough people doing that—but I can see how an unchecked feedback loop and unfettered amplification of positive feedback effects could lead to such a scenario. While caution is warranted, we also need to recognise the tremendous positive impact AI can have on society and its citizens. AI ushers in a new age of hope for many people around the world (World Economic Forum, 2023).

Conclusion

Feedback loops are integral to understanding and improving systems across various domains, from education and software development to artificial intelligence. Action research and agile methodologies exemplify how iterative cycles of planning, feedback, and adjustment foster continuous improvement and innovation.

Feedback loops are not confined to human-designed systems but are also fundamental in nature. In biological systems, feedback mechanisms regulate critical functions such as temperature and glucose levels, ensuring homeostasis. Similarly, in technology, feedback loops in AI and machine learning enable models to adapt and refine their predictions based on continuous input.

Recognising these pervasive feedback loops in both natural and artificial systems highlight their critical role in driving progress and adaptation, underscoring their universal importance in achieving stability and fostering innovation.

And finally, we cannot ignore them, they are everywhere.

References

Adams, F. (2023). Maximizing feedback: The key feedback loops in machine learning lifecycle systems. https://medium.com/@FrankAdams7/maximizing-feedback-the-key-feedback-loops-in-machine-learning-lifecycle-systems-a457b2c9ccc7

Breewood, H. (2018). Feedback loops: What they are and why they matter. https://www.theprogressmotive.org/feedback-loops-what-they-are-and-why-they-matter/

Cockburn, A. (2002). Agile software development. Addison-Wesley.

El-Samad, H. (2021). Biological feedback control—Respect the loops. Cell Systems, 12(6), 477-487. https://doi.org/10.1016/j.cels.2021.05.004

Frakes, R., & Wisher, W. H. (1991). The Terminator. Bantam Books.

Horowitz, P., & Hill, W. (1980). The art of electronics (Chapter 3). Cambridge University Press.

Martin, R. C. (2003). Agile software development: Principles, patterns, and practices. Prentice Hall.

McGregor, D. (2022). Feedback loops build a smarter plant: Data integrity is critical to achieving the right outcomes. Machine Design. https://www.machinedesign.com/automation-iiot

McNiff, J. (2013). Action research: Principles and practice (3rd ed.). Routledge. https://doi.org/10.4324/9780203112755

Parsons, R., & Brown, K. (2002). Teacher as reflective practitioner and action researcher. Wadsworth/Thomson.

Smith, D. (2015). Agile feedback loops. https://medium.com/rootpath/agile-feedback-loops-by-danny-smith-64f6f14894bc

Wakim, S., & Grewal, M. (2020). Human biology. Butte College/ASCC OERI.

World Economic Forum. (2023). Emerging tech like AI are poised to make healthcare more accurate, accessible, and sustainable. https://www.weforum.org/agenda/2023/06/emerging-tech-like-ai-are-poised-to-make-healthcare-more-accurate-accessible-and-sustainable/