Personal agent feeding
"my taxes" application through an API
A reasoner
Personal data
Some rules
A query
Problem statement
Personal agent feeding
"my taxes" application through an API
A reasoner
Personal data
Some rules
A query
Problem statement
People's activities are varied
So it makes sense that their personal agents are configured differently
Can agents themselves help to reach appropriate personal configurations?.
Related work
Semantic Web Services
Services described with the help of agreed semantics
Usually implemented using registries
Allows complex composition of services and views
can be tailored to the user context
Related work
Hypermedia driven APIs
Functionality exposed by means of hypermedia controls
Clients can discover the interfaces in the Web
Allows personalizing content, tailoring the APIs (perhaps according to the data).
Approach
One node (personal agent) can expose Hypermedia APIs generated from:
Personal and public data
Functionality packages and metadata
Approach
Multiple nodes can:
Expose and collect packages from others
Evolve their APIs independently
Approach
While API interfaces evolve, nodes leave evolution trails in the Web, used to:
Discover of new functionality
Discover new relations between the data
Approach
Expectation
Nodes can tailor their functionality, adapting it to the data of a user
The effort of building Hypermedia Interfaces is distributed among peers
API evolution trails are then useful for bootstrapping new configurations
Research Question
The central question investigated by this thesis is:
How to support a process in which independent agents evolve interfaces adapted to the data of
the user, while interoperability between nodes is not sacrificed?
Research Questions
This research raises two concrete questions:
How can API functionality be transfered between nodes in the Web, while still be adapted to the data of
the
user?
How can evolution trails be represented so they are suitable for package discovery and adoption?
Previously defining: What means that an API is adapted to the data of the user?
Hypothesis
Personal agents that exchange API functionality in the Web can build APIs better adapted to the data in
control of
the users,
compared with top down design of application specific API used in client-server architectures
Evaluation plan
Case studies
Personal Agents applied to:
Healthcare
E-government
Evaluation plan
Multi-agent simulation
Knowledge packages scattered in a network
Agents can collect artifacts from others, trying to fulfill goals (or utility function)
Agents leave API evolution trails
Agents can use API evolution trails for discovery
Evaluation plan
Outcomes
Demonstration of feature exposition by means of Hypermedia
Demonstration of feature discovery by means of API evolution trails
Analysis of API articulation efforts, compared to semantic web services
Preliminary results
GPS4IC (Personal) Agent
Hypermedia APIs and reasoning
Transferable Knowledge packages, personal data-spaces
Planify Healthcare paths
Reflections
Bottom-up Approaches
Can be useful to build APIs that adapt to data in control of the user.
Complementary to other approaches.
Has other benefits (does not require giving away the data)
A personal agent-based approach for API evolution
Support slides
At some point (70'), people went from mainframes to personal computers
