<!DOCTYPE html>
<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
</head>
<body>
<b>Study the impact of natural strategies in decision problems</b><br>
<br>
Keywords: Formal Verification, Model Checking for Multi-Agent
Systems, Game Theory<br>
<br>
Description<br>
<br>
Game theory plays a crucial role in AI by providing a mathematical
framework for reasoning about reactive systems, which are defined by
interactions between multiple entities, or players. It has found
significant applications across various fields, including economics,
biology, and computer science. An important application of game
theory in computer science and, more recently, in AI, concerns
formal-system verification. Model checking, introduced in the late
1970s, uses game theory to verify the behavior of systems against
formal specifications. While initial applications focused on closed
systems, which are entirely determined by their internal states,
most systems in practice are open, involving ongoing interactions.
This led to the extension of model checking to Multi-Agent Systems,
using logics like Alternating-time Temporal Logic and Strategy Logic
to account for strategic reasoning.<br>
<br>
Another decision problem in verification is synthesis, the process
of creating a system that meets its specifications across various
environments, particularly those involving rational agents. Rational
synthesis, introduced in the 1990s, addresses this by ensuring that
a system's behavior aligns with specified objectives across
different agent interactions. However, decision problems in MAS
verification can range from polynomial to undecidable, with
complexity heavily influenced by the type of strategy used
(memoryless vs. memoryful). Memoryless strategies are
computationally efficient but less expressive, while memoryful
strategies are more powerful but more complex. To address these
issues, bounded approaches like natural strategies, which consider
simpler strategies in line with bounded rationality, have been
introduced.<br>
<br>
Goals<br>
<br>
The aim of this project is divided into two main steps:<br>
- Study decision problems for natural strategies and analyze
their computational complexity.<br>
- Develop a tool capable of providing answers to decision
problems on natural strategies.<br>
<br>
Profile and skills required<br>
<br>
- PhD in computer science, mathematics, or related fields.<br>
- Strong computer science and/or mathematical background (with
particular attention on formal methods and logic).<br>
- Good programming skills.<br>
- Good level in written and spoken English.<br>
<br>
How to apply<br>
If you are interested you can apply via:
<a class="moz-txt-link-freetext" href="https://institutminestelecom.recruitee.com/l/en/o/post-doctorante-ou-post-doctorant-en-verification-de-systemes-multi-agents">https://institutminestelecom.recruitee.com/l/en/o/post-doctorante-ou-post-doctorant-en-verification-de-systemes-multi-agents</a><br>
<br>
Other information<br>
Application deadline: 31/03/2025<br>
Job type: 18 months fixed-term contract in the ANR NOGGINS project<br>
Job description: <a class="moz-txt-link-freetext" href="https://partage.imt.fr/index.php/s/69zC4zs8PAt5ZNk">https://partage.imt.fr/index.php/s/69zC4zs8PAt5ZNk</a><br>
Scientific contact person: Vadim Malvone
(<a class="moz-txt-link-abbreviated" href="mailto:vadim.malvone@telecom-paris.fr">vadim.malvone@telecom-paris.fr</a>)<br>
</body>
</html>