Human-Centered AI
@human-centered-ai.bsky.social
Focused on understanding and improving how humans interact with artificial intelligence (AI) systems.
Reposted by Human-Centered AI
9/ Finally, #SB205 mandates disclosure when consumers interact with #AI systems.
The simplicity of this requirement belies its profound implications: trust, autonomy, and informed consent in #humanAIinteraction.
The simplicity of this requirement belies its profound implications: trust, autonomy, and informed consent in #humanAIinteraction.
January 18, 2025 at 6:23 PM
9/ Finally, #SB205 mandates disclosure when consumers interact with #AI systems.
The simplicity of this requirement belies its profound implications: trust, autonomy, and informed consent in #humanAIinteraction.
The simplicity of this requirement belies its profound implications: trust, autonomy, and informed consent in #humanAIinteraction.
Reposted by Human-Centered AI
You might like this review and framework paper: link.springer.com/article/10.1...
Using AI to Support Education for Collective Intelligence - International Journal of Artificial Intelligence in Education
Artificial Intelligence (AI) is increasingly being used in education, but mostly to support individual teaching and learning. However, there are good reasons to think that thinking together and solvin...
link.springer.com
January 1, 2025 at 4:55 PM
You might like this review and framework paper: link.springer.com/article/10.1...
9/ Future Perspectives
The paper predicts the need for advancements in IML interfaces, driven by transparency, ethical rigor, and integration with emerging technologies like augmented reality and natural language processing, paving the way for more accessible and impactful #human-AI collaboration.
The paper predicts the need for advancements in IML interfaces, driven by transparency, ethical rigor, and integration with emerging technologies like augmented reality and natural language processing, paving the way for more accessible and impactful #human-AI collaboration.
December 10, 2024 at 3:38 PM
9/ Future Perspectives
The paper predicts the need for advancements in IML interfaces, driven by transparency, ethical rigor, and integration with emerging technologies like augmented reality and natural language processing, paving the way for more accessible and impactful #human-AI collaboration.
The paper predicts the need for advancements in IML interfaces, driven by transparency, ethical rigor, and integration with emerging technologies like augmented reality and natural language processing, paving the way for more accessible and impactful #human-AI collaboration.
8.2/ Foundational Need - Evaluation Metrics
Development of standardized metrics to assess IML system performance, user satisfaction, and ethical compliance is necessary for consistent benchmarking and improvement.
Development of standardized metrics to assess IML system performance, user satisfaction, and ethical compliance is necessary for consistent benchmarking and improvement.
December 10, 2024 at 3:38 PM
8.2/ Foundational Need - Evaluation Metrics
Development of standardized metrics to assess IML system performance, user satisfaction, and ethical compliance is necessary for consistent benchmarking and improvement.
Development of standardized metrics to assess IML system performance, user satisfaction, and ethical compliance is necessary for consistent benchmarking and improvement.
8.1/ Foundational Need - Interdisciplinary Approaches
The paper calls for collaborative efforts across data science, ethics, design, and domain-specific expertise to create holistic solutions that address technical, social, and ethical dimensions
The paper calls for collaborative efforts across data science, ethics, design, and domain-specific expertise to create holistic solutions that address technical, social, and ethical dimensions
December 10, 2024 at 3:38 PM
8.1/ Foundational Need - Interdisciplinary Approaches
The paper calls for collaborative efforts across data science, ethics, design, and domain-specific expertise to create holistic solutions that address technical, social, and ethical dimensions
The paper calls for collaborative efforts across data science, ethics, design, and domain-specific expertise to create holistic solutions that address technical, social, and ethical dimensions
8/ Foundational Need - User-Centered Design Principles
The paper advocates for #usability, #accessibility, and user satisfaction in IML system design. It highlights the need for clear workflows, intuitive visualizations, and inclusivity to ensure the technology benefits a broad user base.
The paper advocates for #usability, #accessibility, and user satisfaction in IML system design. It highlights the need for clear workflows, intuitive visualizations, and inclusivity to ensure the technology benefits a broad user base.
December 10, 2024 at 3:38 PM
8/ Foundational Need - User-Centered Design Principles
The paper advocates for #usability, #accessibility, and user satisfaction in IML system design. It highlights the need for clear workflows, intuitive visualizations, and inclusivity to ensure the technology benefits a broad user base.
The paper advocates for #usability, #accessibility, and user satisfaction in IML system design. It highlights the need for clear workflows, intuitive visualizations, and inclusivity to ensure the technology benefits a broad user base.
7/ An Optimistic Study
A healthcare study demonstrates the potential of IML systems, achieving a 20% reduction in diagnostic time and a 15% increase in accuracy, illustrating their impact in real-world applications.
A healthcare study demonstrates the potential of IML systems, achieving a 20% reduction in diagnostic time and a 15% increase in accuracy, illustrating their impact in real-world applications.
December 10, 2024 at 3:38 PM
7/ An Optimistic Study
A healthcare study demonstrates the potential of IML systems, achieving a 20% reduction in diagnostic time and a 15% increase in accuracy, illustrating their impact in real-world applications.
A healthcare study demonstrates the potential of IML systems, achieving a 20% reduction in diagnostic time and a 15% increase in accuracy, illustrating their impact in real-world applications.
6.6. Risks & Challenges - Model Interpretability
As AI models grow more complex, understanding their decision-making processes becomes increasingly difficult, creating risks of misinterpretation and reduced trust. <--- hello #AISafety
As AI models grow more complex, understanding their decision-making processes becomes increasingly difficult, creating risks of misinterpretation and reduced trust. <--- hello #AISafety
December 10, 2024 at 3:38 PM
6.6. Risks & Challenges - Model Interpretability
As AI models grow more complex, understanding their decision-making processes becomes increasingly difficult, creating risks of misinterpretation and reduced trust. <--- hello #AISafety
As AI models grow more complex, understanding their decision-making processes becomes increasingly difficult, creating risks of misinterpretation and reduced trust. <--- hello #AISafety
6.5/ Risks & Challenges - User Expertise Gaps
IML systems often assume a certain level of user knowledge about machine learning, which can exclude non-technical users and limit the democratization of AI.
IML systems often assume a certain level of user knowledge about machine learning, which can exclude non-technical users and limit the democratization of AI.
December 10, 2024 at 3:38 PM
6.5/ Risks & Challenges - User Expertise Gaps
IML systems often assume a certain level of user knowledge about machine learning, which can exclude non-technical users and limit the democratization of AI.
IML systems often assume a certain level of user knowledge about machine learning, which can exclude non-technical users and limit the democratization of AI.
6.4/ Risks & Challenges - Complexity of Integration
Incorporating IML into existing workflows can be resource-intensive and technically challenging, potentially hindering adoption in organizations with established processes.
Incorporating IML into existing workflows can be resource-intensive and technically challenging, potentially hindering adoption in organizations with established processes.
December 10, 2024 at 3:38 PM
6.4/ Risks & Challenges - Complexity of Integration
Incorporating IML into existing workflows can be resource-intensive and technically challenging, potentially hindering adoption in organizations with established processes.
Incorporating IML into existing workflows can be resource-intensive and technically challenging, potentially hindering adoption in organizations with established processes.
6.3/ Risks & Challenges - Ethical Dilemmas
Decisions made by IML systems, particularly in high-stakes domains like healthcare or finance, may introduce ethical conflicts. Addressing fairness, transparency, and accountability is must be baseline principles.
Decisions made by IML systems, particularly in high-stakes domains like healthcare or finance, may introduce ethical conflicts. Addressing fairness, transparency, and accountability is must be baseline principles.
December 10, 2024 at 3:38 PM
6.3/ Risks & Challenges - Ethical Dilemmas
Decisions made by IML systems, particularly in high-stakes domains like healthcare or finance, may introduce ethical conflicts. Addressing fairness, transparency, and accountability is must be baseline principles.
Decisions made by IML systems, particularly in high-stakes domains like healthcare or finance, may introduce ethical conflicts. Addressing fairness, transparency, and accountability is must be baseline principles.
6.2/ Risks & Challenges - Privacy and Security Concerns
The use of sensitive data for training and real-time interaction raises significant privacy risks. Safeguarding user data and ensuring compliance with privacy regulations are significant real-world challenges.
The use of sensitive data for training and real-time interaction raises significant privacy risks. Safeguarding user data and ensuring compliance with privacy regulations are significant real-world challenges.
December 10, 2024 at 3:38 PM
6.2/ Risks & Challenges - Privacy and Security Concerns
The use of sensitive data for training and real-time interaction raises significant privacy risks. Safeguarding user data and ensuring compliance with privacy regulations are significant real-world challenges.
The use of sensitive data for training and real-time interaction raises significant privacy risks. Safeguarding user data and ensuring compliance with privacy regulations are significant real-world challenges.
6.1/ Risks & Challenges - Bias Amplification
IML systems can inherit or amplify biases present in training data, which may lead to unfair or unethical outcomes. This underscores the need for vigilant bias detection and mitigation.
IML systems can inherit or amplify biases present in training data, which may lead to unfair or unethical outcomes. This underscores the need for vigilant bias detection and mitigation.
December 10, 2024 at 3:38 PM
6.1/ Risks & Challenges - Bias Amplification
IML systems can inherit or amplify biases present in training data, which may lead to unfair or unethical outcomes. This underscores the need for vigilant bias detection and mitigation.
IML systems can inherit or amplify biases present in training data, which may lead to unfair or unethical outcomes. This underscores the need for vigilant bias detection and mitigation.
6/ Risks & Challenges
The following risks highlight the importance of developing ethical guidelines, robust safeguards, and user-friendly interfaces to ensure IML systems are effective, fair, and secure. #AISafety
The following risks highlight the importance of developing ethical guidelines, robust safeguards, and user-friendly interfaces to ensure IML systems are effective, fair, and secure. #AISafety
December 10, 2024 at 3:38 PM
6/ Risks & Challenges
The following risks highlight the importance of developing ethical guidelines, robust safeguards, and user-friendly interfaces to ensure IML systems are effective, fair, and secure. #AISafety
The following risks highlight the importance of developing ethical guidelines, robust safeguards, and user-friendly interfaces to ensure IML systems are effective, fair, and secure. #AISafety
5/ Collaborative frameworks in IML, including HITL, HOTL, HUTL, and Federated Learning, ensure that human involvement is precisely aligned with the complexity and requirements of specific tasks. Sounds dreamy, but there are significant risks with IML...
December 10, 2024 at 3:38 PM
5/ Collaborative frameworks in IML, including HITL, HOTL, HUTL, and Federated Learning, ensure that human involvement is precisely aligned with the complexity and requirements of specific tasks. Sounds dreamy, but there are significant risks with IML...
4/ The accessibility of machine learning is advancing through interfaces like visual dashboards, conversational systems, and gesture-based controls, enabling a broader range of users to collaborate with AI.
December 10, 2024 at 3:38 PM
4/ The accessibility of machine learning is advancing through interfaces like visual dashboards, conversational systems, and gesture-based controls, enabling a broader range of users to collaborate with AI.
3/ IML intends to address challenges such as bias, privacy, and system complexity while delivering significant benefits, including improved accuracy, faster decision-making, and meaningful user engagement.
December 10, 2024 at 3:38 PM
3/ IML intends to address challenges such as bias, privacy, and system complexity while delivering significant benefits, including improved accuracy, faster decision-making, and meaningful user engagement.
2/ IML frameworks like Human-in-the-Loop (HITL) and Human-on-the-Loop (HOTL) emphasize the critical role of humans in actively training or supervising AI systems, enhancing both performance and adaptability.
December 10, 2024 at 3:38 PM
2/ IML frameworks like Human-in-the-Loop (HITL) and Human-on-the-Loop (HOTL) emphasize the critical role of humans in actively training or supervising AI systems, enhancing both performance and adaptability.
1/ Frameworks
• Human-in-the-Loop HITL: Active human involvement in training
• Human-on-the-Loop HOTL: Supervisory roles for decision oversight
• Human-under-the-Loop HUTL: Passive monitoring to ensure compliance
• Federated Learning: Decentralized data model training for privacy-sensitive apps
• Human-in-the-Loop HITL: Active human involvement in training
• Human-on-the-Loop HOTL: Supervisory roles for decision oversight
• Human-under-the-Loop HUTL: Passive monitoring to ensure compliance
• Federated Learning: Decentralized data model training for privacy-sensitive apps
December 10, 2024 at 3:38 PM
1/ Frameworks
• Human-in-the-Loop HITL: Active human involvement in training
• Human-on-the-Loop HOTL: Supervisory roles for decision oversight
• Human-under-the-Loop HUTL: Passive monitoring to ensure compliance
• Federated Learning: Decentralized data model training for privacy-sensitive apps
• Human-in-the-Loop HITL: Active human involvement in training
• Human-on-the-Loop HOTL: Supervisory roles for decision oversight
• Human-under-the-Loop HUTL: Passive monitoring to ensure compliance
• Federated Learning: Decentralized data model training for privacy-sensitive apps
8/ The Takeaway
Designing human-AI systems for the real world isn’t about who does what better. It’s about enabling collective intelligence—humans and machines working together in ways that neither could achieve alone.
The Why is understood
The What is clear
The How this is achieved is TBD
Designing human-AI systems for the real world isn’t about who does what better. It’s about enabling collective intelligence—humans and machines working together in ways that neither could achieve alone.
The Why is understood
The What is clear
The How this is achieved is TBD
December 10, 2024 at 1:52 AM
8/ The Takeaway
Designing human-AI systems for the real world isn’t about who does what better. It’s about enabling collective intelligence—humans and machines working together in ways that neither could achieve alone.
The Why is understood
The What is clear
The How this is achieved is TBD
Designing human-AI systems for the real world isn’t about who does what better. It’s about enabling collective intelligence—humans and machines working together in ways that neither could achieve alone.
The Why is understood
The What is clear
The How this is achieved is TBD
7/ We Need More Research
• Developing frameworks that dynamically integrate evolving AI capabilities.
• Studying how humans and AI truly collaborate in distributed, high-stakes scenarios.
• Understanding how spontaneous coordination emerges in these systems.
• Developing frameworks that dynamically integrate evolving AI capabilities.
• Studying how humans and AI truly collaborate in distributed, high-stakes scenarios.
• Understanding how spontaneous coordination emerges in these systems.
December 10, 2024 at 1:52 AM
7/ We Need More Research
• Developing frameworks that dynamically integrate evolving AI capabilities.
• Studying how humans and AI truly collaborate in distributed, high-stakes scenarios.
• Understanding how spontaneous coordination emerges in these systems.
• Developing frameworks that dynamically integrate evolving AI capabilities.
• Studying how humans and AI truly collaborate in distributed, high-stakes scenarios.
• Understanding how spontaneous coordination emerges in these systems.