Peer | Methodology

Imagine an AI that learns from you, understands your needs, and works alongside you like a true partner. That's the vision behind PEER, merging social science and AI to redefine decision-making. We're trying to go beyond traditional AI to create an AI that's truly useful and trustworthy.

Why the social science connection?

It's all about understanding human needs, values, and expectations. By integrating this understanding into AI development, PEER:

Empowers end-users
Tracks trust and acceptance
Delivers real-world usefulness

PEER achieves this through a cross-disciplinary approach:

Co-creation workshops: These workshops bring together users and AI researchers to engage in collaborative design and discussion, ensuring user needs are reflected in the AI's development.
Wizard-of-Oz prototyping: This technique simulates human-AI interaction, allowing users to experience the AI and provide feedback before formal deployment.
Pilot experimentation: By offering users direct interaction with early versions of the AI, PEER gathers valuable insights and iteratively improves the system based on user feedback.

Why is this important?

The challenge of limited user trust and acceptance remains a critical hurdle in deploying sequential decision-making AI systems. To address this, PEER focuses on co-creation, facilitating deeper bidirectional interaction between humans and AI.

This fosters:

Enhanced mutual learning and reasoning: By directly engaging with the AI, users become active participants in shaping its development and capabilities.
More productive collaborative work: The co-creation process allows for seamless knowledge exchange and collaborative problem-solving, leading to superior outcomes.
Increased user trust and acceptance: Through active participation and transparency, users develop a deeper understanding of AI capabilities and limitations, fostering trust.

Components

Co-creation workshops
Explainable AI
Natural-language, graphical, and interactive interfaces
Design framework for human-centric AI methods
Evaluation and assessment framework for human-centric AI systems
Human-centric AI methods for sequential decision-making

SMIT

Co-creation workshops

The Bigger Picture

The co-creation workshops in PEER aim to involve users and stakeholders in defining human-AI collaborations at the outset of AI development. This process allows participants to identify social, ethical, and technical requirements to enhance interactions between end-users and AI systems.

Through discussions and the use of translations like paper prototypes or tech cards, participants co-create and define their preferred human-AI collaboration, tailored to their specific needs. By simulating these collaborations before development, the project aims to ensure that proposed solutions align with user expectations.

Building the Foundation

Workshops convene users and domain experts early on, leveraging the latest AI advancements to define ideal human-AI collaboration models.
Participants actively shape the project by identifying and evaluating preferred human-AI collaboration configurations based on real-world needs and technical feasibility.

Defining User-Centric AI

Beyond functionality, workshops delve into social, ethical, and technical considerations for optimal user interaction with AI systems.
This collaborative approach ensures PEER develops AI solutions that align with societal values and responsible technology practices.

Translating Algorithms into Action

Complex algorithms underpinning AI are transformed into readily understandable forms like paper prototypes or tech cards.
This user-friendly approach empowers participants to actively co-design and co-create the desired human-AI collaboration for specific use cases.

Benefits of Early Collaboration

User-driven development: Workshops ensure AI systems cater to actual user needs and foster a sense of ownership from the outset.
Continuous feedback loop: Early user input allows for ongoing adjustments and refinements, guaranteeing solutions aligned with user expectations.
Simulating real-world scenarios: By prototyping human-AI collaboration before technical development, researchers can evaluate effectiveness and ensure user-centric functionality.

Tue, CU, JU, AI Lab

Human-centric AI methods for sequential decision-making

Imagine AI that doesn't just answer your questions, but chats with you about the best answer. That's what PEER's human-centered AI is all about: making joint decisions with you, especially for complicated ones.

Where AI Falls

Current AI often explains its decisions once, like a "black box." But for teamwork on complex tasks, this isn't enough. Users can feel left out, unsure why and how decisions are made.

The PEER Difference

We're building AI that learns alongside you:

Two-way conversations: You and AI discuss possible answers, exploring options and adapting to your needs.
Personalized decisions: AI understands your unique preferences and tailors’ solutions just for you.
Clear explanations: You get continuous insights into AI's thought process, building trust and confidence.

This goes beyond simple explanations: Imagine AI helping you plan a trip, considering your budget and interests, or collaborating on a project in real-time, adjusting to your feedback.

The Result

Better decisions: Combine human judgment with AI's strengths for optimal outcomes.
Increased trust: Understandable AI fosters collaboration and empowers you to make informed choices.
Empowerment: AI complements your skills, not replaces them, putting you in control.

CATIE

Evaluation and assessment framework for human-centric AI systems

Building Trustworthy AI: Putting the User in Focus

AI technology is rapidly advancing, but ensuring its trustworthiness is crucial. While researchers focus on ethical principles for development and deployment, the human element – trust and acceptance – is often overlooked.

This project addresses this gap by

Going Beyond Transparency: We recognize the difference between acceptability (pre-use judgment) and acceptance (post-use judgment and behavior).
Measuring What Matters: We're developing reliable metrics to gauge your acceptance, trust, and perception of fairness towards AI systems.
Building the AI Acceptance Index (AIA Index): This innovative tool will assess key factors like interactivity, explainability, and trustworthiness to benchmark the user experience of AI systems like ours.

Our project empowers users by ensuring

Clear Explanations: You'll understand the reasoning behind AI decisions.
Informed Interactions: You'll have the tools to interact with AI systems confidently.
Trustworthy Technology: You'll be able to rely on AI to make informed decisions.

Tue, CU, JU, AI Lab

Explainable AI

Artificial intelligence (AI) has become increasingly prevalent across various domains, from healthcare to retail, finance, and beyond. However, as AI systems play a more significant role in decision-making processes that impact our lives, issues around user trust and acceptance have emerged.

One reason for this is the lack of understanding and transparency in AI decision-making, leading to user mistrust. Additionally, current AI systems often fail to comprehend user needs and preferences, resulting in solutions that are not adequately personalized or tailored to individual situations. As a result, users may feel that AI systems hinder rather than empower them, reducing their autonomy.

PEER's Solution

We're developing AI that engages in deeper, ongoing interactions with humans. This means:

Transparency: You'll understand how AI decisions are made, building trust and confidence.
Two-way learning: AI continuously learns from you, leading to personalized solutions that adapt to your needs.
Collaboration: Humans and AI work together as partners, tackling problems more effectively.

This goes beyond explaining single decisions. Imagine AI that learns your driving preferences or collaborates with you on managing a project in real-time.

The Impact

Better Decisions: Combine human judgment with AI's analytical power for optimal results.
Increased Trust: Understandable AI leads to better collaboration and user confidence.
Empowered Users: AI complements your skills, not replaces them, leaving you in control.

Int, EUT

Natural-language, graphical,
and interactive interfaces

Empowering Users in Sequential Decision-Making with Natural Language Processing, Interactive Visualizations, and Adaptive Explanations.

Traditional Explainable AI (XAI) systems often struggle with sequential decision-making tasks, offering rigid options and limited explanations. PEER breaks this mold by integrating a novel Multimodal XAI framework that fosters dynamic user interaction and trust.

Core Technologies

Natural Language Processing (NLP): Users can interact with PEER conversationally, using plain language for a more intuitive and user-friendly experience.
Interactive Visualizations: Complex concepts are presented through clear and informative charts and graphs, enhancing clarity of explanations during sequential decision processes.
Adaptive Explanations: The system leverages an interactive interface, allowing users to tailor explanations to their specific needs, preferences, and learning styles.

Benefits of PEER's Multimodal XAI

Transparency by Design: PEER prioritizes transparency, offering users a complete understanding of the reasoning behind each recommendation.
Customization for Individual Needs: The interactive interface allows explanations to be customized for each user, fostering a more user-centric approach.
Building Trust Through Interaction: By actively engaging with users and adapting explanations, PEER builds trust in its recommendations.

PEER Newsletter

Subscribe to our newsletter to get the latest news and updates.

Methodology

Social science meets AI