Dear all,
We will have our next LIRa session tomorrow, on Wednesday, 26 April 16:30.
Note the unusual day: Because Thursday is a national holiday in The Netherlands we will have our session on Wednesday instead.
This will be an ONLINE-ONLY session. To attend, please use our recurring zoom link:
https://uva-live.zoom.us/j/89230639823?pwd=YWJuSnJmTDhXcWhmd1ZkeG5zb0o5UT09
(Meeting ID: 892 3063 9823, Passcode: 421723)
You can find the details of the talk below.
Speaker: Jon Williamson (University of Kent)
Date and Time: Wednesday, April 26th 2023, 16:30-18:00
Note the unusual day! (Due to Thursday being a holiday.)
Venue: ONLINE only (i.e. not hybrid)
Title: A decidable class of inferences in first-order objective Bayesian inductive logic
Abstract. This talk begins with a gentle introduction to objective
Bayesian inductive logic and the main differences between this
approach and Carnap's approach to inductive logic. Then we move to the
question of decidability. First-order deductive logic is well known to
be undecidable and this undecidability carries over to inductive logic
under the standard semantics, which holds that premisses inductively
entail a conclusion just when every probability function that
satisfies the premisses also satisfies the conclusion. It is perhaps
surprising then that a large class of inferences in objective Bayesian
inductive logic turns out to be decidable. We see that decidability is
achieved by reducing the general inference problem to a problem
involving only quantifier-free premisses, which can be solved using a
truth-table method. This reduction permits the use of Bayesian
networks to provide a potentially efficient approach to inference in
objective Bayesian inductive logic.
Joint work with Juergen Landes & Soroush Rafiee Rad.
Hope to see you there!
The LIRa team
Dear all,
We will have our next LIRa session on Wednesday, 26 April 16:30.
Note the unusual day: Because Thursday is a national holiday in The Netherlands we will have our session on Wednesday instead.
This will be an ONLINE-ONLY session. To attend, please use our recurring zoom link:
https://uva-live.zoom.us/j/89230639823?pwd=YWJuSnJmTDhXcWhmd1ZkeG5zb0o5UT09
(Meeting ID: 892 3063 9823, Passcode: 421723)
You can find the details of the talk below.
Speaker: Jon Williamson (University of Kent)
Date and Time: Wednesday, April 26th 2023, 16:30-18:00
Note the unusual day! (Due to Thursday being a holiday.)
Venue: ONLINE only (i.e. not hybrid)
Title: A decidable class of inferences in first-order objective Bayesian inductive logic
Abstract. This talk begins with a gentle introduction to objective
Bayesian inductive logic and the main differences between this
approach and Carnap's approach to inductive logic. Then we move to the
question of decidability. First-order deductive logic is well known to
be undecidable and this undecidability carries over to inductive logic
under the standard semantics, which holds that premisses inductively
entail a conclusion just when every probability function that
satisfies the premisses also satisfies the conclusion. It is perhaps
surprising then that a large class of inferences in objective Bayesian
inductive logic turns out to be decidable. We see that decidability is
achieved by reducing the general inference problem to a problem
involving only quantifier-free premisses, which can be solved using a
truth-table method. This reduction permits the use of Bayesian
networks to provide a potentially efficient approach to inference in
objective Bayesian inductive logic.
Joint work with Juergen Landes & Soroush Rafiee Rad.
Hope to see you there!
The LIRa team
Dear all,
Tomorrow at 16:30 we will have a joint LLAMA and LIRa session!
This will be a *hybrid* session. If you want to attend online, please
use the following ZOOM LINK, DIFFERENT THAN OUR USUAL ONE:
https://uva-live.zoom.us/j/84413536364
Meeting ID: 844-1353-6364
You can find the details of the talk below.
This will be a joint session together with the LLAMA seminar.
Speakers: Ekaterina Kubyshkina & Mattia Petrolo
Date and Time: Thursday, April 20th 2023, 16:30-18:00
Venue: ILLC seminar room F1.15 in Science Park 107 and online.
Title: Creating new worlds: a PAL for excusable ignorance.
Abstract. There is a lively debate in the current literature on
epistemology on which type of ignorance may provide a moral excuse. A
good candidate is the one in which an agent has never thought about or
considered as true a proposition p. From a logical perspective, no
formal analysis was provided for ignorance as an excuse. We fill this
gap by providing an original logical setting for modelling this type
of ignorance. In particular, we introduce a complete and sound logic
in which ignorance is expressed as a primitive modality. Semantically,
the logic is characterized by Kripke semantics with possibly
incomplete worlds, or, equivalently, Kripke models equipped with a
three-valued valuation function. Moreover, to consider the conditions
of a possible change of an agent's ignorance, we will extend the
setting to public announcement logic. First, we show that standard
update procedures are not suitable for an adequate representation of
scenarios involving excusable ignorance. Secondly, we define an
original update procedure, which modifies the initial models by
creating new worlds and accessibilities, instead of eliminating them.
We exemplify the fruitfulness of this procedure via an intuitive
example.
(joint work with M. K. Pereira)
Hope to see you there!
The LIRa team
Dear all,
We will have a joint LLAMA and LIRa session next week on Thursday, 20 April 16:30.
This will be a hybrid session. If you want to attend online, please
use the following ZOOM LINK, DIFFERENT THAN OUR USUAL ONE:
https://uva-live.zoom.us/j/84413536364
Meeting ID: 844-1353-6364
You can find the details of the talk below.
Speakers: Ekaterina Kubyshkina & Mattia Petrolo
Date and Time: Thursday, April 20th 2023, 16:30-18:00
Venue: ILLC seminar room F1.15 in Science Park 107 and online.
Title: Creating new worlds: a PAL for excusable ignorance.
Abstract. There is a lively debate in the current literature on
epistemology on which type of ignorance may provide a moral excuse. A
good candidate is the one in which an agent has never thought about or
considered as true a proposition p. From a logical perspective, no
formal analysis was provided for ignorance as an excuse. We fill this
gap by providing an original logical setting for modelling this type
of ignorance. In particular, we introduce a complete and sound logic
in which ignorance is expressed as a primitive modality. Semantically,
the logic is characterized by Kripke semantics with possibly
incomplete worlds, or, equivalently, Kripke models equipped with a
three-valued valuation function. Moreover, to consider the conditions
of a possible change of an agent's ignorance, we will extend the
setting to public announcement logic. First, we show that standard
update procedures are not suitable for an adequate representation of
scenarios involving excusable ignorance. Secondly, we define an
original update procedure, which modifies the initial models by
creating new worlds and accessibilities, instead of eliminating them.
We exemplify the fruitfulness of this procedure via an intuitive
example.
(joint work with M. K. Pereira)
Hope to see you there!
The LIRa team
Dear all,
Unfortunately the LIRa session tomorrow is cancelled and there will be no LIRa this week.
Our next event will be a joint session with the LLAMA seminar next week, on Thursday, April 20th.
kind regards,
The LIRa team
Dear all,
We will have our next LIRa session tomorrow, on Thursday, 6 April 16:30.
This will be an ONLINE ONLY session. To attend please use our recurring zoom link:
https://uva-live.zoom.us/j/89230639823?pwd=YWJuSnJmTDhXcWhmd1ZkeG5zb0o5UT09
(Meeting ID: 892 3063 9823, Passcode: 421723)
You can find the details of the talk below.
Speaker: Emiliano Lorini
Date and Time: Thursday, April 6th 2023, 16:30-18:00
Venue: ONLINE ONLY. (i.e. not hybrid!)
Title: Non-Classical Logics for Explanations in AI Systems
Abstract. I will show how non-classical logics with special emphasis
on modal logic, epistemic logic, conditional logic and deontic logic
can be used to represent and compare a rich variety of explanations in
classifier systems widely used in the field machine learning including
abductive, constrastive, counterfactual, objective vs subjective,
interactive as well as concepts of bias, discrimination and protected
feature. I will present proof-theoretic and complexity results for
these logics and illustrate their expressiveness through some concrete
examples.
Hope to see you there!
The LIRa team