Re: Agentic Linked Data -- attempt 2

Hello Melvin and Joshua,
I'm interested. To continue the discussion (which I supposed should
eventually get its own topical home in a more Linked Data-specific
forum), I think FOAF has nicely sorted out the senses of "Agent". FOAF has
an Agent as both a super class of Person, but also is the class of anything
that "does stuff" <http://xmlns.com/foaf/spec/#term_Agent>. So I think
the concept of a machine-based Agent is fair game, and we can use FOAF or
other ontologies to sort of what exact type of agent it is (what are all of
a specific Agent instance's classes, in addition to Agent).

As far as Agentic AI or any sense of the term Agent, I think it starts
boiling down to the Agent's contextual role, and that role entails
permissions and responsibilities.

Okay, I'll stop there, but I'm excited to see where this goes. As an aside,
one thing Henry Story and I chatted about a while ago, in relation to
Solid, was the use of Linked Data to represent household relationships for
Health and Human services Linked Data. Households, household memberships,
roles (and agents with roles), etc. were all critical to the concept of
client consent and fine-grained information sharing in this vast domain.

--

Eric Jahn
CTO/Data Architect
Alexandria Consulting LLC
St. Petersburg, Florida
727.537.9474
alexandriaconsulting.com
WebID <https://alexandriaconsulting.com/files/eric_jahn.rdf#me>

On Sat, Dec 28, 2024 at 6:51 AM Joshua Cornejo <josh@marketdata.md> wrote:

> Hi Melvin,
>
> I would be interested.
>
> A lot of the work I’ve done in the authorisation architecture with ODRL
> has been driven by both “agents” (as potential delegates) and embedding
> security (e.g. different pieces of information generate the same embedding,
> and now you need to segregate/obfuscate for the linked paths).
>
> A piece of advice is to define what "Agent " means, in OIDC user accounts
> are assigned to humans and machine accounts (service accounts) are assigned
> to machines, following slightly different protocols.
>
> From what I see in the examples of “agents” have mixed both with a machine
> used for connection to a service provider (API keys = service accounts)
> whilst the “personas” within the implementation of an agent try to emulate
> humans. I haven’t done much thinking about it - but “smells” wrong.
>
> Regards,
> _________________________
> *Joshua Cornejo*
> *marketdata <https://www.marketdata.md/>*
> smart authorisation management for the AI-era
>
> On 28 Dec 2024, at 11:11, Melvin Carvalho <melvincarvalho@gmail.com>
> wrote:
>
> Hi All,
>
> When I started this group, it was with the idea that Solid = Social Linked
> Data. Solid's foundation on WebID ties it to the concept of "Agent," based
> on FOAF's Agent class (parent of FOAF Person).
>
> With "agentic AI" really taking off right now, I wonder—would anyone be
> interested in exploring "Agentic Linked Data"? It feels like a natural
> direction.
>
> Best,
>
> Melvin
>
>
>