Intelligent Software Agents (II)

Nick Morakitis wrote a nice paper in the heydays of the  Intelligent Software Agents: “Intelligent Software Agents Application and Classification“. And what is better it comes under the section “Where Can Intelligent Agents be Used in Travel Reservation Systems?” of the Surprise 97 conferences in the Imperial College of Science Technology, (May-June 1997).

Nick Morakitis starts saying : The evolution of modern software agent technology has given rise to an extensive overuse of the term agent. The result of this abusive usage of the word agent and the terminology that relates to it, is a great confusion as to what an agent is, what types of agents can be identified, and what are the properties and characteristics with respect to which the aforementioned identifications can be achieved.

Nick makes a summary of the up to date notions of Software Agents, describing the weak and strong definitions of Agents, the classification by their application. The explosive growth of the field was creating endless applications and uses for any subject: Personal Assistants/Organizers, Preferential Information Systems, Financial Decisions, Entertainment, Military, Transport ..

According to his classification, Zentric’s aim of making the Internet more reachable to Users saving much of their time falls into the category of  Interface+Information Agents.  They would be … “Agents focused in autonomy and learning and whose goal is to perform tasks for their owner are called interface agents. Typically, such agents observe the actions of the user, interact with him/her and attempt to learn by imitation, or direct feedback. Their abilities in negotiation are limited and most of their interactions with other agents (if any) relate to receiving advisory information. Interface agents are seemingly smart software pieces that help their owner automate and perform tasks…  and … Information agents are tools that coordinate the retrieval, management, selection and manipulation of information.”

In most of the literature related to IA (Intelligent Agents) I notice the influence of AI (Artificial Intelligence) and its highly aspirational academic style. Great theories about intelligent software , self learning, adaptive to the user, etc. etc. These are deep waters when it comes to creating software that must work and succeed beyond university labs. 

I would drop the idea of building intelligent software. Considering that nobody knows what intelligence is, we’ll have to admit that the term Intelligent Agents is more an aspirational/motivational term, it’s some way of saying that we’re are trying to build something smart, something that would make the User more efficient and his interaction with the software more satisfactory, something that as the relationship advances the efficiency and satisfaction increases thanks to the adaptation of the software to the User.

The next question is whether it would be preferable to build the Zentric Software using the principle of Intelligent Agents. To answer this question I need to find out if there are SDK or platforms that would fit me. So I’ll have to check JADE from FIPA, COUGAAR and the framework of  Carnegie Mellon University.

Intelligent Agents. Nice. Where are they ? (I)

Oh! I’m very late, this quest it’s an old aspiration. I hit the “Intelligent Agents” (IA). Plenty of stuff on Google.

Goal driven, task oriented, smart, mobile, intercomunicated processes, self learners, Agents working for me, exactly. We need lots of them to create the UCI, an intelligent agent to find jobs, another one to find trip deals and so on.

Reading papers mainly of the 90′s  it looked like Intelligent Agents were going to take the Internet in the near future. Beautiful concept and even more beautiful software architecture to implement them, a dream system for programers. They were very exciting years for development of theories and software. I think it was a general enthusiasm in all fields around Internet, many of them were ahead of their time and maybe IA was one of them. By what I see in their websites most of them are discontinued, strangely enough most of them were shading off around 1998. 

So far I have found out three relevant currently active efforts on the field, 1. The Foundation of Physical Agent (FIPA),  2. COUGAAR an Open Source Architecture for large scale distributed multi-agent systems and 3. the impressive work of Katia Sycara at the The Intelligent Software Agents Lab of the Carnegie Mellon University.

Three of them are working for a long time now and have solutions for creating intelligent distributed agent systems. They have many answers to my questions, specially Katia Sycara and her team, have analized , proposed and developed software that could be considered the ground of the User Centric Internet (UCI). 

In the paper  “A Taxonomy of Middle-agents for the Internet,” Katia Sycara and Chi Wong propose the model for the Users or “Requesters” and Providers (let’s say current websites) to communicate through the Middle Agent (MA). This model fits with my view of the UCI.

In another paper called LARKS for short, Katia Sycara and colleges go much further to define a key issue: How to describe an Agent Capability for agents to understand each other of what they offer and therefore other agents know what they and how can they ask for it. 

COUGAAR is extremely interesting for the solutions they offer and the understanding the show of large scale systems as it would be our case in the UCI. We can see their knowledge and experience when talking about the Yellow Pages implementation techniques for the listing of providers.

Three of the case studies I’m talking have their open source to implement multi-agent solutions.