=Paper=
{{Paper
|id=Vol-3618/pe_paper_2
|storemode=property
|title=PAIRS - Privacy-aware, intelligent and resilient crisis management
|pdfUrl=https://ceur-ws.org/Vol-3618/pe_paper_2.pdf
|volume=Vol-3618
|authors=Sabine Janzen,Agbodzea P. Ahiagble,Lotfy Abdel Khaliq,Natalie Gdanitz,Prajvi Saxena,Prathvish Mithare,Denys Skrytskyi,Wolfgang Maass
|dblpUrl=https://dblp.org/rec/conf/er/JanzenAKGSMS023
}}
==PAIRS - Privacy-aware, intelligent and resilient crisis management==
https://ceur-ws.org/Vol-3618/pe_paper_2.pdf
PAIRS – Privacy-aware, intelligent and resilient crisis
management
Sabine Janzen1,∗ , Agbodzea P. Ahiagble1 , Lotfy Abdel Khaliq1 , Natalie Gdanitz1 ,
Prajvi Saxena1 , Prathvish Mithare1 , Denys Skrytskyi1 and Wolfgang Maass1,2
1
Deutsches Forschungszentrum für Künstliche Intelligenz (DFKI), Saarbruecken, Germany
2
Saarland University (UdS), Saarbrücken, Germany
Abstract
In an era characterized by unpredictability and increasingly interconnected systems, crisis management
becomes a formidable challenge. This paper introduces the PAIRS project (06/2021 - 05/2024), an
ongoing research initiative funded by the German Federal Ministry of Economics and Climate Protection,
aiming to enhance crisis management through AI-based smart crisis management services. Building
on artificial intelligence (AI) and large-scale data, PAIRS empowers a shift from reactive to proactive
crisis management in organizations spanning multiple sectors, including civil protection, healthcare,
production & supply chains, and energy. The PAIRS consortium, a multidisciplinary collaboration of
eleven industry and academic partners, explores the use of smart crisis management services to effectively
manage the complexity and scale of modern crises, ensuring fast recovery and improved resilience. This
paper provides a comprehensive overview of the PAIRS project, outlining its objectives, work packages,
and anticipated outcomes. A central focus is the role of conceptual modeling from both process and
product perspectives within PAIRS, with specific emphasis on the current state of work in modeling
crisis scenarios across various domains.
Keywords
Crisis management, Artificial Intelligence, Resilience, Smart Crisis Management Services, Conceptual
modeling, Crisis scenarios, Knowledge graph
1. Introduction
In an increasingly interconnected and dynamic world, crises are inevitable, hard to predict,
and pose immense challenges that demand robust, effective and timely management. Over
the last decade, various types of crises have impacted societies, economies, and environments
globally [1]. For instance, the COVID-19 pandemic in 2019 unleashed a global health crisis with
sweeping socio-economic consequences [2, 3, 4]. Further, in 2020, the wildfires that devastated
Australia and California underlined the urgency of dealing with climate crises [5]. Meanwhile,
the 2021 Suez Canal obstruction demonstrated the vulnerability of global supply chains [6].
ER2023: Companion Proceedings of the 42nd International Conference on Conceptual Modeling: ER Forum, 7th SCME,
Project Exhibitions, Posters and Demos, and Doctoral Consortium, November 06-09, 2023, Lisbon, Portugal
∗
Corresponding author.
Envelope-Open sabine.janzen@dfki.de (S. Janzen); agbodzea_pascal.ahiagble@dfki.de (A. P. Ahiagble);
lotfy.abdel_khaliq@dfki.de (L. A. Khaliq); natalie.gdanitz@dfki.de (N. Gdanitz); prajvi.saxena@dfki.de (P. Saxena);
prathvish.mithare@dfki.de (P. Mithare); denys.skrytskyi@dfki.de (D. Skrytskyi); wolfgang.maass@dfki.de
(W. Maass)
© 2023 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR
Workshop
Proceedings
http://ceur-ws.org
ISSN 1613-0073
CEUR Workshop Proceedings (CEUR-WS.org)
CEUR
ceur-ws.org
Workshop ISSN 1613-0073
Proceedings
While these crises differ in their nature and scale, they all require rapid response and effective
management. Traditional methods, often plagued by human bias and slow response times, can
fall short of successfully managing these crises. Artificial Intelligence (AI) has created new
opportunities to improve crisis management leveraging machine learning, predictive analytics,
and natural language processing, among other technologies, to collect, analyze, and interpret
vast amounts of data in real-time [7]. This enables the identification of emerging crises, the
prediction of their potential impacts, the creation of response strategies, and the implementation
of those strategies in a timely and effective manner.
PAIRS is an ongoing research project funded by the German Federal Ministry of Economics
and Climate Protection (06/2021 - 05/2024) that investigates privacy-aware, intelligent and
resilient crisis management1 . By means of AI and data-driven smart crisis management
services (SCMS), companies, government and health organizations as well as civil protection are
enabled to move from a reactive stance to a more proactive one, anticipating crises and mitigating
their impacts. The project consortium consists of eleven partners from industry and academia
that take different roles in the project: industry end users (Sick, Bisping), civil protection end
users (THW), system providers (Advaneo, Tiplu, IBM), and research & development (DFKI,
Fraunhofer IPA, FIR at RWTH Aachen, Saarland University, OFFIS). PAIRS investigates the
application of AI-based SCMS in the domains of civil protection, production & supply chains,
healthcare, and energy with the objective to handle the complexity and scale of contemporary
crises, ensuring quicker recovery and better resilience.
In this paper, we present an overview of the PAIRS project, including its objectives, work
packages and expected outcomes. We will discuss the role of conceptual modeling in PAIRS
from a process and product perspective, especially focusing on the current state of work in
modeling crisis scenarios in diverse domains.
2. Project objectives
PAIRS aims to develop a cross-domain platform for crisis management to identify and anticipate
crisis scenarios on the basis of a hybrid AI approach. In doing so, the availability of essential
resources and capabilities of enterprise ecosystems shall be secured, and their marketability
sustainably strengthened. In particular, the interactions between economic and political actors
will be considered, taking into account data sovereignty and data privacy. The risks of individual
crisis scenarios (concrete manifestations of the crisis event together with the general reactions)
are assessed on an actor-specific basis. Based on the risk assessment and a self-expanding pool
of measures, individual response measures are derived for individual crisis scenarios, which are
fed back (pseudo-)anonymized into a crisis scenario generator. The resulting iterative precision
of the crisis scenarios, together with the continuous adaptation of all response measures,
contributes to a well-founded information basis in order to be prepared in the best possible
way for potential crisis situations. The project’s work packages have been distributed among
the partners based on their respective areas of expertise. As depicted in Figure 1, the project
is structured into ten sub-projects, each labeled as a work package (WP). WP1, led by FIR at
RWTH Aachen, aims to define, structure, and prioritize crisis management use cases using a
1
https://www.dfki.de/en/web/research/projects-and-publications/project/pairs
knowledge base. WP2 is led by ADVANEO and focuses on investigating and consolidating
existing reference architectures while mapping general requirements. ADVANEO also leads
WP3, tasked with developing data integration strategies for PAIRS-compliant connection of
services and data governance. WP4, under the leadership of DFKI, involves analyzing existing
research approaches, AI algorithms, modules, frameworks, and tools for core technical topics.
Saarland University takes charge of WP5, which involves evaluating the feasibility of extensive
data anonymization for AI tools in PAIRS from functional and legal perspectives. Fraunhofer
IPA leads WP6, responsible for researching and identifying relevant resilience approaches for
stable production and supply chains in crisis situations, as well as seizing business opportunities.
WP7, led by FIR, focuses on the conceptual design of the procedure for acceptance and benefit
validation. ADVANEO also leads WP8, responsible for the conceptual design of acceptance and
benefit validation procedures, integration of partners’ exploitation paths, and defining economic
goals. Finally, Saarland University leads WP9, which focuses on creating an ecosystem for
project exploitation during and beyond the project.
Figure 1: List of the work packages in PAIRS project (Source: PAIRS project).
3. Conceptual modeling for smart crisis management services
(SCMS)
Conceptual modeling has a pivotal role in PAIRS, bridging theoretical constructs with practical
AI-based solutions for crisis management and was applied from a product and process perspec-
tive within the project. In this section, we demonstrate the role of conceptual modeling within
work package 4 (WP4), i.e., process perspective, as well as regarding first results with respect to
use cases and services, i.e., product perspective.
3.1. Work package 4: Process perspective
The goal of WP4 is the conceptual and technical specification as well as prototypical imple-
mentation of specific and generic AI modules for SCMS in iterative proof-of-concept (PoC)
cycles. In the foundational phase, a comprehensive state analysis was conducted to review
the current research strategies, AI algorithms, AI modules offered by system providers, and
relevant tools and frameworks. Following this, relevant data sets were identified, acquired, and
processed. Next, implementation strategies for core technical subjects such as explainable and
responsible AI, AI-based crisis communication and episodic crisis scenario knowledge graphs
were developed. Followed by technical specifications of the required AI modules, AI models
were designed, trained and integrated into modules for the PAIRS platform. Last, benchmarking
of the performance of these modules is done in an evaluation phase. In order to optimally
address the challenging technical issues in WP4, DFKI, as WP leader, developed a proposal
for combining the horizontally designed work packages with a vertical agile logic. For this
purpose, the SCRUM approach is adapted to research projects following the SCORE method [8].
Five sprint categories are distinguished: Knowledge Transfer, Research, Conceptual Modelling,
Implementation, and Publication, as shown in Figure 2. Each project month is organized as
a sprint of one category. At the beginning of each month, a sprint jour fixe (JF) takes place
in WP4, where the last sprint is reviewed and the next sprint is planned. In each Knowledge
Figure 2: Sprint categories and work planning in PAIRS work package 4 (TP4) according to SCORE
method.
Transfer sprint, the team determines essential knowledge that needs to be shared for effective
collaboration. Knowledge Transfer sprints take place in project months 6, 19, 28 and 33 (match-
ing the milestones of PAIRS). Research sprints cover knowledge engineering work, i.e., research
and analysis of existing work, approaches, algorithms, tools, etc. Conceptual Modelling sprints
aim at the development of new solution concepts and the specification of conceptual models
for these solutions informed by earlier sprint outcomes. Implementation sprints focus on the
technical realization of pre-defined concepts, starting with the alignment of user stories for
the upcoming PoC or prototype. This sprint might also include evaluations or studies beyond
pure technical developments. Publication sprints induce the objective of converting findings
of the previous sprint into results, i.e., writing scientific publications, (technical) report, white
paper, state of the arts, etc. Planning and executing the sprints in WP4 are aligned with the
overall project goals by the work package lead (”Are the intended results of the current sprint
aligned with the WP and project goals?”). Furthermore, sprints are documented in condensed
form (sprint plan, user stories, sprint results, review). Five sprint cycles were completed in the
project until September 2023. An overview of the AI modules that are planned, in progress or
completed and thus available can be found in Figure 3.
Figure 3: Overview of AI modules that are work-in-progress (wip), planned (open) or done in PAIRS
work package 4: Smart Crisis Management Services (SCMS) (09/2023). Modules have titles and involved
respective responsible partners in PAIRS. Modules can be applied in single or multiple use cases defined
in PAIRS: civil protection, energy, supply chain and production as well as healthcare. After integration of
modules in use case-specific SCMS prototypes are available that can be used as PAIRS demonstrators.
3.2. AI modules and use cases: Product perspective
In PAIRS WP4, several modules were developed and evolved to use case-specific service
implementations (product perspective) (cf. Figure 3). Resulting services were showcased within
publications, demonstrators, one pagers and screencasts. An extract of finalized modules will
be highlighted in this section.
Outage predictor - Prediction of regional power outages for industrial produc-
tion: Power outages and fluctuations represent serious crisis situations in energy-intensive
process industries (e.g., glass and paper production), where substances such as oil, gas,
wood fibers or chemicals are processed. Power disruptions can interrupt chemical reactions
and produce tons of waste as well as damage of machine parts [9, 10, 11]. But, despite of
the obvious criticality, handling of outages in manufacturing focuses on commissioning
of expensive proprietary power plants to protect against power outages and implicit gut
feeling in anticipating potential disruptions [9]. Within this use case, three modules (i.e.,
AI-based scenario planning (AISOP), scenario patterns, outage predictor) were aggregated to a
service prototype (Python, JSON-LD, Neo4j, Django, HTML, JS) for predicting energy-driven
disruptions (cf. Figure 3). We introduced AISOP (cf. Figure 4) as a model for AI-based scenario
Figure 4: AI-based scenario planning for predicting crisis situations: AISOP module.
planning for predicting crisis situations that uses conceptual, well-defined scenario patterns
(JSON-LD) to capture entities of crisis situations within an scenario knowledge graph that
can be used to predicting future crisis scenarios by predictive analytics [9]. The model was
exemplified within an outage predictor service2 that enables the prediction of regional power
outages for locations of the German paper industry for max. 7 days (accuracy: 0.81, sensitivity:
0.70)[12]3 .
Hidden problem detector - Identification of hidden problems in supply chains:
Component-based supply chains of products are increasingly non-transparent beyond tier 1
suppliers [13, 14]. Disruptions in early stages remain undetected, propagate, and reinforce
before popping up as critical situations at tier 1 [15, 16, 17]. Despite criticality, traditional
supply chain management focuses on reactive measures at tier 1 or 2 for these hidden problems
[18]. In PAIRS GRASPER was developed (cf. Figure 5), a model for graph-theoretical analysis
of component criticality that uses multiple centrality measures to detect hidden problems
in component-based supply chains [17]. Bill-of-Materials (BOM) data are automatically
Figure 5: Model for graph-theoretic analysis of component criticality in supply chains: Hidden problem
detector module.
2
https://github.com/InformationServiceSystems/pairs-project/tree/main/Modules/OutagePredictor
3
https://www.youtube.com/watch?v=pj6K4pOvoDs&t=18s
mapped onto a knowledge graph, semantically enriched, and fed with historical and actual
market data (e.g., availabilities, prices) provided by Octopart 4 [17]. The Criticality Controller
performs graph-theoretic analysis with respect to in-degree, out-degree, out-strength, and
betweenness-centrality of nodes to determine the criticality of supply chains on a component
level. The model is applied in the Hidden Problem Detector (Flask, Python, Neo4j, Octopart)5
for hidden problem detection in sensor manufacturing supply chains (cf. Figure 3) [17]6 .
Social signal observer - Social signal detection for crisis prediction: Crises emit
weak early warning signals, difficult to detect amid daily noise [19, 20, 21]. Signal detection
mechanisms in crisis management aim for early identification and proactive organizational
responses [22, 23, 24, 25]. Observation of social signals in information sources (e.g., social
media) enables early identification of crises supporting proactive organizational responses
before a crisis occurs. In PAIRS, we introduced an AI model to support open-domain signal
detection of crisis-related indicators in Twitter posts that was instantiated in a service Social
Signal Observer (Python, Flask, HTML, CSS, Javascript) (cf. Figure 3) [21]7 . Here, we work
Figure 6: Model for open-domain social signal detection of crisis-related indicators in tweets: Social
signal observer.
with multi-lingual Twitter data and combine multiple state-of-the-art models such as, GPT-3,
RoBERTa 8 , and STANZA [26]. The service accepts inserted keywords by the user regarding
crisis signals of interest along with a selection of country, language and time frame as input.
For detected domain-specific signals, alerts with confidence and severity are triggered and
presented.
4
https://octopart.com/
5
https://github.com/InformationServiceSystems/pairs-project/tree/main/Modules/HDP
6
https://www.youtube.com/watch?v=Xixp7x7hhQU&t=24s
7
https://github.com/InformationServiceSystems/pairs-project/tree/main/Modules/OSOS
8
https://huggingface.co/joeddav/xlm-roberta-large-xnli
4. Conclusion and future work
Facing an ever-evolving global landscape marked by multifaceted crises, the need for advanced,
agile, and efficient crisis management tools has never been more pronounced. The PAIRS
project, as delineated in this paper, investigates the potential of harnessing AI and data-driven
methods for transforming the way we approach, understand, and mitigate crises. By integrating
privacy-aware, intelligent, and resilient strategies, PAIRS not only addresses the immediate
challenges of crisis management but also anticipates future threats, ensuring a proactive stance.
Accompanied by the collaborative and diverse nature of the project consortium consisting of
industrial and academic partners, conceptual modeling plays a crucial role in PAIRS, both from
a process and product perspective, for enabling a structured, low-threshold and result-driven
approach to AI-based crisis management. This ensures that the solutions developed are both
adaptable to diverse domains and scalable to the magnitude of contemporary crises. There are
several avenues for future work including the technical and empirical evaluation of the modules
including studies with decision makers. In conclusion, PAIRS offers a promising glimpse into
the future of crisis management. By leveraging the power of AI, it seeks to transform reactive
responses into proactive strategies to be more resilient by means of smart crisis management
services.
Acknowledgments
This work was partially funded by the German Federal Ministry of Economics and Climate
Protection (BMWK) under the contract 01MK21008D.
References
[1] H. V. D. Ven, Y. Sun, Varieties of crises: Comparing the politics of covid-19 and climate
change, Global Environmental Politics 21 (2021) 13–22. doi:10.1162/glep_a_00590 .
[2] A. Martin, M. Markhvida, S. Hallegatte, B. Walsh, Socio-economic impacts of covid-19 on
household consumption and poverty, Economics of disasters and climate change 4 (2020)
453–479. doi:10.1007/s41885- 020- 00070- 3 .
[3] M. Nicola, Z. Alsafi, C. Sohrabi, A. Kerwan, A. Al-Jabir, C. Iosifidis, M. Agha, R. Aghaf,
The socio-economic implications of the coronavirus pandemic (covid-19): a review, Inter-
national Journal of Surgery 78 (2020) 185–193. doi:10.1016/j.ijsu.2020.04.018 .
[4] M. Alradhawi, N. Shubber, J. Sheppard, Y. Ali, Effects of the covid-19 pandemic on mental
well-being amongst individuals in society- a letter to the editor on “the socio-economic
implications of the coronavirus and covid-19 pandemic: a review, International Journal of
Surgery 78 (2020) 147–148. doi:10.1016/j.ijsu.2020.04.070 .
[5] C. News, California’s disastrous wildfires pale in comparison to australia’s
catastrophic fires, 2020. URL: https://www.cbsnews.com/sanfrancisco/news/
australia-bushfires-california-wildfires-season-comparison/.
[6] J. M. yin Lee, E. Y. cheung Wong, Suez canal blockage: An analysis of legal impact, risks
and liabilities to the global supply chain, MATEC web of conferences 339 (2021) 01019.
doi:10.1051/matecconf/202133901019 .
[7] G. D. Hager, R. Bryant, E. Horvitz, M. Mataric, V. Honavar, Advances in arti-
ficial intelligence require progress across all of computer science, 2017. URL:
https://www.researchgate.net/publication/318460915_Advances_in_Artificial_
Intelligence_Require_Progress_Across_all_of_Computer_Science.
[8] M. Hicks, J. S. Foster, Adapting scrum to managing a research group (2010). URL: https:
//drum.lib.umd.edu/bitstream/1903/10743/1/CS-TR-4966.pdf.
[9] S. Janzen, N. Gdanitz, L. A. Khaliq, T. Munir, C. Franzius, W. Maaß, Anticipating energy-
driven crises in process industry by ai-based scenario planning, HICSS 56/23. Hawaii
International Conference on System Sciences (2023).
[10] E. Ahadu, The effect of electric blackout on the operation and productivity of small
manufacturing enterprises, Journal of Energy Technologies and Policy 9 (2019) 1–11.
doi:10.7176/jetp/9- 6- 01 .
[11] I. Ahmed, M. Baddeley, D. Coffman, J. Meikle, G. Sianjase, The cost of power outages to
zambia’s manufacturing firms, International Growth Centre, London School of Economic
and Political Science (2019). URL: https://discovery.ucl.ac.uk/id/eprint/10134294.
[12] N. Gdanitz, L. H. A. Khaliq, A. P. Ahiagble, S. Janzen, W. Maaß, Powop: Weather-based
power outage prediction, Intelligent Systems Conference (IntelliSys) (2023).
[13] M. Caridi, L. Crippa, A. Perego, A. Sianesi, A. Tumino, Do virtuality and complexity affect
supply chain visibility?, International Journal of Production Economics 127 (2010) 372–383.
doi:10.1016/j.ijpe.2009.08.016 .
[14] T. K. Agrawal, J. O. Ravi Kalaiarasan and, M. Wiktorsson, Supply chain visibility: A
delphi study on managerial perspectives and priorities, International Journal of Production
Research (2022) 1–16. doi:10.1080/00207543.2022.2098873 .
[15] Y. Sheffi, Preparing for disruptions through early detection, MIT Sloan Management
Review 57 (2022) 31–42.
[16] T. Y. Choi, D. Rogers, B. Vakil, Coronavirus is a wake-up call
for supply chain management, 2020. URL: https://hbr.org/2020/03/
coronavirus-is-a-wake-up-call-for-supply-chain-management.
[17] S. Janzen, S. Baer, A. P. Ahiagble, W. Maaß, Tackling non-transparency - identification of
hidden problems in component-based supply chains, 20th International Conference on
Information Systems for Crisis Response and Management (ISCRAM) (2023).
[18] K. Katsaliaki, P. Galetsi, S. Kumari, Supply chain disruptions and resilience: a major
review and future research agenda, Annals of Operations Research 319 (2021) 965–1002.
doi:10.1007/s10479- 020- 03912- 1 .
[19] K.-w. Fu, Y. Zhu, Did the world overlook the media’s early warning of covid-19?, Journal
of Risk Research 23 (2020) 1047–1051.
[20] C. Diks, C. Hommes, J. Wang, Critical slowing down as an early warning signal for
financial crises?, Empirical Economics 57 (2019) 1201–1228.
[21] S. Janzen, P. Saxena, S. Baer, W. Maaß, ”listening in”: Social signal detection for crisis
prediction, in: HICSS 57/24. Hawaii International Conference on System Sciences (HICSS-
2024), Hawaii, United States, HICSS, 2024.
[22] J. Wolbers, S. Kuipers, A. Boin, A systematic review of 20 years of crisis and disaster
research: Trends and progress, Risk, Hazards & Crisis in Public Policy 12 (2021) 374–392.
[23] T. Hensgen, K. C. Desouza, G. D. Kraft, Games, signal detection, and processing in the
context of crisis management, Journal of Contingencies and Crisis Management 11 (2003)
67–77.
[24] S. Elsubbaugh, R. Fildes, M. B. Rose, Preparation for crisis management: A proposed
model and empirical evidence, Journal of contingencies and crisis management 12 (2004)
112–127.
[25] J. A. Parnell, W. R. Crandall, What drives crisis readiness? an assessment of managers
in the united states: The effects of market turbulence, perceived likelihood of a crisis,
small-to medium-sized enterprises and innovative capacity, Journal of Contingencies and
Crisis Management 29 (2021) 416–428.
[26] P. Qi, Y. Zhang, Y. Zhang, J. Bolton, C. Manning, Stanza: A python natural language
processing toolkit for many human languages, 2020, pp. 101–108. doi:10.18653/v1/2020.
acl- demos.14 .