Decision Support with Interactive Advanced Data-Enabled Multiobjective Optimization Systems (DESIDES)
Main funder
Funder's project number: 355346
Funds granted by main funder (€)
- 599 876,00
Funding program
Project timetable
Project start date: 01/09/2023
Project end date: 31/08/2027
Summary
Our mission is to support making better decisions in the presence of conflicting objectives. We support decision makers (DMs) in finding the best balance with advanced interactive multiobjective optimization methods. Our specialty lies in explanations and visualizations. We also develop methods for groups of DMs. Overall, we develop next-generation support for intelligible decisions.
With interactive multiobjective optimization (MOO) methods, a DM iteratively provides preference information to find the most preferred compromise. Current MOO methods do not adequately learn from the preferences or explain tradeoffs. And there are no openly available interactive tools for groups of DMs. Thus, we develop novel interactive MOO methods, tools for group decision making and appropriate visualizations with integrated explainability to support making better decisions.
We advance the state-of-the-art by supporting DMs in understanding complex interdependencies, explaining tradeoffs among objectives and learning a DM’s or DMs’ preferences. We also support DMs with different preferences in finding an acceptable consensus. Our data-enabled methods make the most of data available complemented by domain expertise of DMs. Further novelty lies in developing visualizations to communicate insight. With them, DMs can better comprehend the consequences of decisions and gain confidence in the selected decision.
We inspire, evaluate and validate our research and further improve the methods with use cases in energy system optimization, sustainable forestry and robotics engineering in collaboration with domain experts. All cases have conflicting objectives, e.g. in forestry examples are income, sustainability, recreational values and habitat suitability.
In our next-generation decision support, we incorporate explainability to solution processes of a single or multiple DMs. We i) fully utilize preferences of DMs provided during the interactive decision process and support DMs by personalized decision alternatives and explanations. Thus, DMs can ii) place greater trust in how decisions are derived and iii) justify decisions to stakeholders involved. iv) Novel visualizations facilitate exchange of information and increase understanding. We v) implement the novel methods as elements of a modular open-source software framework DESDEO. Thus, they are available for further developments and for other problem domains since the main elements to be developed are general by nature.
With interactive multiobjective optimization (MOO) methods, a DM iteratively provides preference information to find the most preferred compromise. Current MOO methods do not adequately learn from the preferences or explain tradeoffs. And there are no openly available interactive tools for groups of DMs. Thus, we develop novel interactive MOO methods, tools for group decision making and appropriate visualizations with integrated explainability to support making better decisions.
We advance the state-of-the-art by supporting DMs in understanding complex interdependencies, explaining tradeoffs among objectives and learning a DM’s or DMs’ preferences. We also support DMs with different preferences in finding an acceptable consensus. Our data-enabled methods make the most of data available complemented by domain expertise of DMs. Further novelty lies in developing visualizations to communicate insight. With them, DMs can better comprehend the consequences of decisions and gain confidence in the selected decision.
We inspire, evaluate and validate our research and further improve the methods with use cases in energy system optimization, sustainable forestry and robotics engineering in collaboration with domain experts. All cases have conflicting objectives, e.g. in forestry examples are income, sustainability, recreational values and habitat suitability.
In our next-generation decision support, we incorporate explainability to solution processes of a single or multiple DMs. We i) fully utilize preferences of DMs provided during the interactive decision process and support DMs by personalized decision alternatives and explanations. Thus, DMs can ii) place greater trust in how decisions are derived and iii) justify decisions to stakeholders involved. iv) Novel visualizations facilitate exchange of information and increase understanding. We v) implement the novel methods as elements of a modular open-source software framework DESDEO. Thus, they are available for further developments and for other problem domains since the main elements to be developed are general by nature.
Principal Investigator
Other persons related to this project (JYU)
Primary responsible unit
Follow-up groups
Profiling area: Decision analytics utilizing causal models and multiobjective optimization (University of Jyväskylä JYU) DEMO; 2017-2021
