Unit 10

Last but not least - Module Reflection

This reflection provides a critical overview of my academic and technical development during the Intelligent Agents module. Drawing upon discussions, practical exercises, the team development project, and the challenges of the individual implementation task, I reflect on how my understanding of agent-based computing, intelligent systems design, and AI ethics has evolved. While the module was demanding at times, it ultimately offered meaningful insights into the real-world application of autonomous agents and AI systems.

Discussion Engagement and Analytical Development

The initial discussion on agent-based systems helped establish core motivations for decentralised architectures, such as flexibility, resilience, and real-time adaptability. I explored examples across manufacturing, robotics, and autonomous vehicles, while peer responses expanded the conversation to fault tolerance, semantic interoperability, and explainability. These exchanges clarified how agent models are increasingly tied to both AI and system reliability concerns in complex domains.

The second discussion on agent communication languages allowed me to examine protocols like KQML and FIPA-ACL in relation to both semantic clarity and implementation complexity. My post contrasted these with more efficient, tightly coupled alternatives, and peers introduced hybrid solutions that reinforced the value of layered protocol design. This discussion helped me appreciate the balance between interoperability and performance—particularly important when designing scalable agent systems.

I did not contribute to the third discussion on deep learning ethics, largely due to time constraints and a lack of earlier peer feedback. In hindsight, this was a missed opportunity for engagement. It reminded me that learning is not solely about feedback volume but about developing critical consistency in participation. I aim to maintain a more stable presence in future modules, even when motivation fluctuates.

Practical Activities and Applied Skill-Building

The KQML/KIF agent dialogue task gave me hands-on practice constructing formal agent communication using performative speech acts. It reinforced the value of structured knowledge representation in enabling autonomy, coordination, and meaningful exchange between agents. The constituency-based parse tree task offered insights into how intelligent systems analyse natural language input. It also helped me reconnect with foundational concepts in computational linguistics that are essential for agent-based dialogue systems.

For the deep learning application exercise, I explored deepfakes as a socially impactful use of AI. Analysing their technical foundation (GANs) alongside privacy, ethical, and security concerns sharpened my awareness of how intelligent systems intersect with public trust, governance, and misuse. The exercise reminded me that intelligent systems must be evaluated not only on what they can do, but also on what they ought to do.

Team Project and Real-World System Design

Working with Group E on our digital forensics project, we designed an autonomous verification agent integrated into a Chrome extension. The system intercepts downloads and applies rule-based and machine learning methods to detect and quarantine suspicious files. I contributed to technical research, architecture definition, and modelling the communication and risk assessment pipeline. Our approach combined structured YARA-based filtering with lightweight anomaly detection models to ensure performance and portability.

This project helped translate theoretical knowledge into a tangible, practical solution with real cybersecurity implications. It also strengthened my appreciation of collaborative design, especially in aligning system goals with compliance and ethical standards such as GDPR and ISO 27001. One insight was the importance of intervening early in the security chain—before threats reach endpoints—rather than relying solely on host-based detection systems.

Individual Development and Technical Challenges

The most demanding part of the module was the individual development of the agent prototype, which extended our team’s project. Revisiting Python for this purpose required me to re-learn several libraries and build confidence in file scanning, cryptographic signing, and API communication. I encountered significant challenges in integrating the Chrome extension with the server backend and debugging asynchronous behaviours. These setbacks were frustrating at times, but ultimately they pushed me to improve my technical self-reliance and problem-solving skills.

This experience reinforced the importance of combining theoretical design with practical fluency. It also highlighted areas for future growth, including more advanced Python engineering, testing practices, and greater familiarity with browser-extension development. Despite initial setbacks, delivering a working agent prototype was a rewarding milestone and a meaningful step forward in my applied AI capabilities.

Personal and Professional Growth

Throughout the module, I developed a more comprehensive understanding of agent systems—not only how they function, but also how they fit into broader technological and ethical ecosystems. I strengthened my written and technical communication skills, particularly through detailed documentation and structured analysis. The integration of AI ethics into our project planning also challenged me to think more critically about accountability and transparency in intelligent system design.

While my engagement could have been more consistent in discussion activities, the combination of team collaboration and individual development gave me a holistic experience. This reflection process has clarified my strengths in system design, research integration, and conceptual modelling, while also revealing areas for improvement in real-time coding confidence and sustained academic participation.

Conclusion

In summary, the Intelligent Agents module has significantly enhanced my understanding of multi-agent systems, autonomous design principles, and the ethical dimensions of AI. Through practical activities, critical discussion, and project-based learning, I have gained valuable insights into both the potential and complexity of intelligent systems. Moving forward, I intend to build on this foundation by applying these concepts in professional contexts related to cybersecurity, digital governance, and applied AI innovation.

Most important References

  • Acar, A., Lu, L., Uluagac, A.S. and Kirda, E. (2019) ‘An Analysis of Malware Trends in Enterprise Networks’, arXiv preprint arXiv:1910.00508. Available at: https://arxiv.org/abs/1910.00508.
  • Devlin, J., Chang, M.-W., Lee, K. and Toutanova, K. (2019) ‘BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding’, in Proceedings of NAACL-HLT 2019. Minneapolis, MN: ACL. Available at: https://aclanthology.org/N19-1423/.
  • Labrou, Y. and Finin, T. (1998) ‘Semantics and conversations for an agent communication language’, arXiv preprint cs/9809034. Available at: https://arxiv.org/abs/cs/9809034.
  • Leitão, P., Colombo, A.W. and Karnouskos, S. (2016) ‘Industrial automation based on cyber-physical systems technologies: Prototype implementations and challenges’, Computers in Industry, 81, pp. 11–25. https://doi.org/10.1016/j.compind.2015.08.004.
  • Mayfield, J., Labrou, Y. and Finin, T. (1996) ‘Evaluation of KQML as an agent communication language’, in Wooldridge, M.J., Müller, J.P. and Tambe, M. (eds.) Intelligent Agents II: Agent Theories, Architectures, and Languages. Berlin: Springer, pp. 347–360.
  • Verdoliva, L. (2020) ‘Media Forensics and DeepFakes: An Overview’, IEEE Journal of Selected Topics in Signal Processing. https://ieeexplore.ieee.org/document/9115874.
  • Verizon (2025) 2025 Data Breach Investigations Report. Verizon Enterprise Solutions. Available at: https://www.verizon.com/business/resources/Te7d/reports/2025-dbir-data-breach-investigations-report.pdf.
  • Wolsey, A. (2022) ‘The State-of-the-Art in AI-Based Malware Detection Techniques: A Review’, arXiv preprint arXiv:2210.11239. Available at: https://arxiv.org/abs/2210.11239.