Learning Outcomes

Intelligent Agents

AI generated with Bing
  1. Identify and critically analyse agent-based systems, differentiating between architectures and approaches.
  2. Apply and critically evaluate intelligent agent techniques to real-world problems, particularly where technical risk and uncertainty is involved.
  3. Deploy critically appropriate software tools and skills for the design and implementation of an agent-based system, bearing in mind applicable legal, social, ethical, and professional issues.
  4. Systematically develop and implement the skills required to be effective member of a development team in a virtual professional environment, adopting real-life perspectives on team roles and organisation.

Artefacts

Collaborative Discussion 1

Topic

Agent Based Systems: Discuss what has led to the rise of agent-based systems and the benefits that this approach can offer to organisations.

Learning Outcomes

  • Outcome 1. Discussed agent autonomy, its interaction with an environment using sensors and actuators, and its social ability (in addition to reactivity and proactiveness). Reviewed use in IoT in Industry 4.0 and as a virtual developer in transformer-based technology to generate source code.
  • Outcome 4. Social ability raised philosophical deliberation: is Stephen Hawking intelligent if he cannot communicate? Further debate considered whether intelligent agent use is declining or evolving.

Feedback

Ungraded. Peer response: Argued that social ability was beneficial but not required, and that social capabilities in humans do not define intelligence. This was an insightful debate.

Artefacts

Collaborative Discussion 2

Topic

Agent Communication Languages: Discuss potential advantages and disadvantages of agent communication languages such as KQML and how they compare with method invocation in Python or Java.

Learning Outcomes

  • Outcome 1. Discussed speech act, performative "action verbs". Compared to FIPA-ACL, JADE and SPADE and contrasted object-oriented method invocation with proactive autonomy of agents.
  • Outcome 4. Further deliberation debated relevance of KQML and clarified agent autonomy, as agents cannot invoke each other's methods.

Feedback

Ungraded. Peer response: Asked me to consider value of innovation. Did it drive improvement or was it for glory? Another insightful point of consideration.

Artefacts

Development Team Project: Academic Research Online Agent

Plan

Topic

Develop an agent capable of Academic Research Online: find search results on a website, extract the data, and send to an offline location. Deliver a comprehensive design proposal report.

Learning Outcomes

  • Outcome 1. Created design proposal for Academic Research Online System (AROS) simple reflex agent. It uses Python, requests, Beautiful Soup 4 and Pandas in collect, extract, transform, with a DOM tree-based approach. Modular PRINCE2 Agile planning and iteration enables extensions like Scrapy for crawling, Selenium for automation and NLTK for NLP.
  • Outcome 2. Applied a simple reflex agent to academic research online. Data changes over time and fault tolerance and error handing will be needed beyond prototype to account for volume, variety, velocity, and veracity. Code will need to accommodate both representational and connectivity errors.
  • Outcome 3. As this is an early-knowledge team, requests, Beautiful Soup and Pandas were selected for this design stage over Scrapy. Future iterations need legislation and ethical considerations including robots.txt, rate limiting, not scraping at peak hours, respecting terms of use, and avoiding private data.
  • Outcome 4. The team was virtual across time zones. Maria set up a shared Google Drive, with communication via WhatsApp and Microsoft Teams. Planning and iteration used PRINCE2 Agile. Each team member created an initial outline to gain understanding and growth. Maria's shared interactive outline with design decisions, rationale and references became the base. She merged and completed the first full draft which was reviewed and edited by the team. Incremental progress was shared via WhatsApp.

Tutor Feedback

84% (Distinction) "The work is very good, and the mark should show that!" "Overall this is a good piece of work, well done!" Positives were raised around the well-articulated design decisions, particularly the choice of Python libraries. The diagrams were considered to be used well, the writing classed as well-structured and easy to follow, and credibility was added with citations. Minor points focused on design explanations rather than agent descriptions, and acknowledged that word count limitation restricted further addition.

Artefacts

Development Individual Project: AROS Simple Reflex Agent and Presentation

AROS

Topic

Developed Academic Research Online System (AROS) using a simple reflex agent. AROS finds a search term on arXiv and stores relevant results in a CSV file. Code and README are in GitHub. Presentation (and associated transcript) includes Python code development over time, discussion of error handling, demonstration of PyTest unit testing, and demonstrates code execution, and discusses steps needed for a 1.0 release.

Learning Outcomes

  • Outcome 1. The primary analysis was done in the design phase. Here, however, the original simple prototype script was converted into a class with methods to enhance modularity and maintainability. Hardcoded URLs were parameterised to improve flexibility.
  • Outcome 2. Selected arXiv as it was easily scrapable with clear citation data. AROS handles a range of errors.
  • Outcome 3. Used Python, requests, Beautiful Soup 4, Pandas, with PyTest for unit testing. Future roadmap explains handling larger scale issues.
  • Outcome 4. This is an individual assignment.

Tutor Feedback

83% (Distinction) Described as a "thoughtful approach to code development", "well-communicated...strong understanding of the material", "comprehensive understanding of web scraping techniques", and "promise for continued development". The only request, described as non-essential, was to speak more on error handling.

Formative Activities

Ungraded formative activities / homework by unit.

Learning Outcomes

  • Unit 6: KQML and KIF Agent Dialogue - Outcomes 1, 2
  • Unit 8: Constituency-based Parse Trees - Outcomes 1, 2

Reflections

Critical reflection on learning.

SWOT Analysis

SWOT

Tutor Feedback

83% (Distinction). Overall: "The evidence you have provided is great, you show an in-depth view of the contributions to the module. You demonstrate a strong understanding of the work and a fantastic level of commitment to the learning." On goals: "Your aim of ethically advancing in media and AI, your plan to continuously learn and apply coding, AI, ML, and AI ethics showcases a forward-thinking and responsible mindset." On reflection: "Your reflection is self-aware, and demonstrates continuous learning, and ethtical considerations in professional growth."

Reflection available upon request.