PhD Course on AI-augmented Process Management and Mining

Teachers: Simone Agostinelli & Andrea Marrella

When: From 1/07/2024 to 5/07/2024, Room B203, 9:30-13:30 or online at https://uniroma1.zoom.us/j/89062433567?pwd=bwJFbHNdIAXbvTHBJhkGun89du4vVa.1

Contact email: agostinelli@diag.uniroma1.it

Abstract: State-of-the-art Artificial Intelligence (AI) solutions are increasingly used to improve the automation and accuracy of business processes (BPs), consisting of interrelated tasks executed according to business rules for achieving some enterprise goal, such as producing a service for customers. In this direction, the recent literature on Business Process Management (BPM) has identified the research challenges to developing a new breed of AI-augmented BPM Systems (ABPMSs). While in a conventional BPMS each task or decision in a BP is driven either by a human agent or by a software application according to predetermined business logic, in an ABPMS, the system can reason about the current state of the BP to determine on the fly the course of actions that improves its performance, without predefined execution scripts. Also, the ABPMS must support BPs with an unknown or incomplete structure at design-time that may emerge at run-time as a result of reasoning and learning from past executions, or because of external events produced by the working environment. In this course, we provide an overview of the theoretical foundations and practical implementations of those AI techniques that are contributing to realize a paradigm shift from the reactive conventional BPM philosophy towards the proactive mechanisms required by ABPMSs. Specifically, we show how techniques leveraging temporal logics, situation- and action-based formalisms, model learning, automated planning, and large language models enable the effective management of the main lifecycle phases of an ABPMS. 

Registration: https://forms.gle/G8oevhSBBSaDjUZA9

Shared folder for slides: https://drive.google.com/drive/folders/1LxgKC6gEUL--gqKqm8QcgBsH1NVTCSpo?usp=drive_link

Lectures timeline (passcode for accessing zoom links: AI4PMM2024@PHD):

1/7/2024 AI-augmented Process Management Systems (S. Agostinelli, M. Montali), 4 hours

1. [9:30-11:30] Introduction to the course, Basics of Process Management and Mining, Background on AI-augmented BPM systems (S. Agostinelli): 

2. [11:30-13:30] Process Framing via LTLf techniques (M. Montali), 2 hours

2/7/2024 Predictive Process Mining (P. Bisconti, C. Di Francescomarino/F.M. Maggi), 4 hours

3. [9:30-11:30] What is Trustworthiness made of? Between the AI Act and the perceived trustworthiness (P. Bisconti)

4. [11:30-13:30] Predictive Monitoring of Business Processes (C. Di Francescomarino/F.M. Maggi)

3/7/2024 Automated Planning-as-a-Service for AIBPMSs (A. Marrella), 4 hours

5. [9:30-11:30] Conformance Checking through automated planning in AI

6. [11:30-13:30] Process adaptation using reasoning about actions techniques

4/7/2024 Robotic Process Automation and Mining (S. Agostinelli), 4 hours

7. [9:30-11:30] Robotic process mining, Segmentation of user interface logs 

8. [11:30-13:30] Reactive synthesis of software robots from user interface logs

5/7/2024 Process Management in the era of Big Data (F. Leotta, M.L. Bernardi), 4 hours

9. [9:30-11:30] Context-aware and IoT-based process intelligence (F. Leotta)

10. [11:30-13:30] Conversational AI for BPM with Large Language Models (M.L. Bernardi)

Total Hours: 20

Exam: Two alternatives are available to the students to pass this exam:

• Paper presentation. Students present the content of 2/3 research papers assigned by the teacher. No groups are allowed.

• Implementation of a small-scale project accompanied by a discussion of the outcomes, whether positive or negative, pertinent to the student's research domain. Groups are allowed.

Estimated workload of 5-6 days.