Skip to main content

Resultados relacionados

Resultados totales (Incluyendo duplicados): 4
Encontrada(s) 1 página(s)

UPCommons. Portal del coneixement obert de la UPC
Publication Artículo científico (JournalArticle)

Assessing biases through visual contexts

  • Arias Duart, Anna|||0000-0002-8819-6735
  • Giménez Ábalos, Víctor
  • Cortés García, Claudio Ulises|||0000-0003-0192-3096
  • Garcia Gasulla, Dario|||0000-0001-6732-5641
Bias detection in the computer vision field is a necessary task, to achieve fair models. These biases are usually due to undesirable correlations present in the data and learned by the model. Although explainability can be a way to gain insights into model behavior, reviewing explanations is not straightforward. This work proposes a methodology to analyze the model biases without using explainability. By doing so, we reduce the potential noise arising from explainability methods, and we minimize human noise during the analysis of explanations. The proposed methodology combines images of the original distribution with images of potential context biases and analyzes the effect produced in the model’s output. For this work, we first presented and released three new datasets generated by diffusion models. Next, we used the proposed methodology to analyze the context impact on the model’s prediction. Finally, we verified the reliability of the proposed methodology and the consistency of its results. We hope this tool will help practitioners to detect and mitigate potential biases, allowing them to obtain more reliable models., This work received funding from the European Union’s H2020-INFRAIA-2019-1 program under the Grant Agreement n.871042 (SoBigData++: European Integrated Infrastructure for Social Mining and Big Data Analytics) and from the HORIZON-INFRA-2021-DEV-02 program under the Grant Agreement n.101079043 (SoBigData RI Preparatory Phase Project). Additionally, this work was supported by the Departament de Recerca i Universitats of the Generalitat de Catalunya, under the Industrial Doctorate Grant DI 2018-100., Peer Reviewed
UPCommons. Portal del coneixement obert de la UPC
Publication Contribución de congreso (ConferenceOutput)

Policy graphs and intention: answering 'why' and 'how' from a telic perspective

  • Giménez Ábalos, Víctor
  • Álvarez Napagao, Sergio|||0000-0001-9946-9703
  • Tormos Llorente, Adrián
  • Cortés García, Claudio Ulises|||0000-0003-0192-3096
  • Vázquez Salceda, Javier|||0000-0003-1732-9446
Agents are a special kind of AI-based software in that they interact in complex environments and have increased potential for emergent behaviour. Explaining such behaviour is key to deploying trustworthy AI, but the increasing complexity and opaque nature of many agent implementations makes this hard. In this work, we reuse the Policy Graphs method --which can be used to explain opaque agent behaviour-- and enhance it to query it with hypotheses of desirable situations. These hypotheses are used to compute a numerical value to examine agent intentions at any particular moment, as a function of how likely the agent is to bring about a hypothesised desirable situation. We emphasise the relevance of how this approach has full epistemic traceability, and each belief used by the algorithms providing answers is backed by specific facts from its construction process. We show the numeric approach provides a robust and intuitive way to provide telic explainability (explaining current actions from the perspective of bringing about situations), and allows to evaluate the interpretability of behaviour of the agent based on the explanations; and it opens the door to explainability that is useful not only to the human, but to an agent., This work has been partially supported by the AI4EUROPE (Grant agreement ID: 101070000), SoBigData PPP (Grant agreement ID: 101079043) and V. Giménez-Abalos fellowship within the “Generación D” initiative, Red.es, Ministerio para la Transformación Digital y de la Función Pública, for talent atraction (C005/24-ED CV1). Funded by the European Union NextGenerationEU funds, through PRTR., Peer Reviewed
Recercat. Dipósit de la Recerca de Catalunya
Publication Artículo científico (JournalArticle)

Assessing biases through visual contexts

  • Arias Duart, Anna
  • Giménez Ábalos, Víctor
  • Cortés García, Claudio Ulises
  • Garcia Gasulla, Dario
Bias detection in the computer vision field is a necessary task, to achieve fair models. These biases are usually due to undesirable correlations present in the data and learned by the model. Although explainability can be a way to gain insights into model behavior, reviewing explanations is not straightforward. This work proposes a methodology to analyze the model biases without using explainability. By doing so, we reduce the potential noise arising from explainability methods, and we minimize human noise during the analysis of explanations. The proposed methodology combines images of the original distribution with images of potential context biases and analyzes the effect produced in the model’s output. For this work, we first presented and released three new datasets generated by diffusion models. Next, we used the proposed methodology to analyze the context impact on the model’s prediction. Finally, we verified the reliability of the proposed methodology and the consistency of its results. We hope this tool will help practitioners to detect and mitigate potential biases, allowing them to obtain more reliable models., This work received funding from the European Union’s H2020-INFRAIA-2019-1 program under the Grant Agreement n.871042 (SoBigData++: European Integrated Infrastructure for Social Mining and Big Data Analytics) and from the HORIZON-INFRA-2021-DEV-02 program under the Grant Agreement n.101079043 (SoBigData RI Preparatory Phase Project). Additionally, this work was supported by the Departament de Recerca i Universitats of the Generalitat de Catalunya, under the Industrial Doctorate Grant DI 2018-100., Peer Reviewed, Postprint (published version)
Recercat. Dipósit de la Recerca de Catalunya
Publication Contribución de congreso (ConferenceOutput)

Policy graphs and intention: answering 'why' and 'how' from a telic perspective

  • Giménez Ábalos, Víctor
  • Álvarez Napagao, Sergio
  • Tormos Llorente, Adrián
  • Cortés García, Claudio Ulises
  • Vázquez Salceda, Javier
Agents are a special kind of AI-based software in that they interact in complex environments and have increased potential for emergent behaviour. Explaining such behaviour is key to deploying trustworthy AI, but the increasing complexity and opaque nature of many agent implementations makes this hard. In this work, we reuse the Policy Graphs method --which can be used to explain opaque agent behaviour-- and enhance it to query it with hypotheses of desirable situations. These hypotheses are used to compute a numerical value to examine agent intentions at any particular moment, as a function of how likely the agent is to bring about a hypothesised desirable situation. We emphasise the relevance of how this approach has full epistemic traceability, and each belief used by the algorithms providing answers is backed by specific facts from its construction process. We show the numeric approach provides a robust and intuitive way to provide telic explainability (explaining current actions from the perspective of bringing about situations), and allows to evaluate the interpretability of behaviour of the agent based on the explanations; and it opens the door to explainability that is useful not only to the human, but to an agent., This work has been partially supported by the AI4EUROPE (Grant agreement ID: 101070000), SoBigData PPP (Grant agreement ID: 101079043) and V. Giménez-Abalos fellowship within the “Generación D” initiative, Red.es, Ministerio para la Transformación Digital y de la Función Pública, for talent atraction (C005/24-ED CV1). Funded by the European Union NextGenerationEU funds, through PRTR., Peer Reviewed, Postprint (published version)