I was recently recognized by #MichiganAI, #UMichCSE as an #AI #RisingStar.
I put forth an agenda of #CollaborativeHumanAISystems - #IntelligentSystems that are designed to model, reason, and learn about their human partners -supporting them in their goals and pursuits.
It is #AI, #ML research but extremely #HumanCentered. We start with studying what is the human trying to do and what intelligent support do they need.
A thread below -
In our intro to #AI we have been told that this is what an #AI #System, an #Agent is.
But this is not the correct picture.
In any reasonable deployment, there is going to be humans involved.
Humans are the MAIN decision makers in an #AI #System deployment. We can't ignore them. We can't pretend they don't exist. We can't pretend that thinking about the humans doesn't matter and that that #autonomy is everything.
#CollaborativeHumanAISystems studies 3 questions:
1. How do we model a human so that an #AI #System can reason with it? How to find a _prescriptive_ model? Human-centered sciences have a starting point.
2. How do we design an #intelligent #collaborator #agent?
3. How to we measure progress? THE most critical question. As #CS people, we are enamored byaccuracy, scale etc. But, in #HumanCentered research, we have to study what humans care about.
The potential #impact is interdisciplinary.
#CollaborativeHumanAISystems 1: #HealthBehaviorChange, #AICoaching for #Health, #CognitiveAI
AAAI 2017: https://ojs.aaai.org/index.php/AAAI/article/view/19101
ACM TiiS 2020: https://arxiv.org/abs/1910.04836
ACM TiiS 2021: https://arxiv.org/abs/1910.07728
JMIR 2017: https://www.jmir.org/2017/11/e397/
IEEE EMBC 2016: https://ieeexplore.ieee.org/abstract/document/7591428
#CollaborativeHumanAISystems 2: #Sustainability #Transportation #BehaviorInfluence, #CognitiveAI
#JAIR 2019: https://www.jair.org/index.php/jair/article/view/11352
#AIES 2019: https://dl.acm.org/doi/abs/10.1145/3306618.3314271
#ACM #IUI #Humanize 2019:
#CollaborativeHumanAISystems 3: #ITL, #InteractiveTaskLearning, #HumanRobotInteraction - how can complex #agents learn from humans? #language is large part of that puzzle.
Focus on #ELP, #EmbodiedLanguageProcessing - what does it mean to 'understand' language for communication, collaboration, & teaching.
Not #InformationRetrieval #IR which #NLP research studies.
#IEEE #ROMAN 2021: https://arxiv.org/abs/2102.06755
#ACS 2020: https://arxiv.org/abs/2006.01962
#ACS 2014: https://arxiv.org/abs/1604.02509
Interactive Task Learning (ITL) is an emerging research agenda that studies the design of complex intelligent robots that can acquire new knowledge through natural human teacher-robot learner interactions. ITL methods are particularly useful for designing intelligent robots whose behavior can be adapted by humans collaborating with them. Various research communities are contributing methods for ITL and a large subset of this research is \emph{robot-centered} with a focus on developing algorithms that can learn online, quickly. This paper studies the ITL problem from a \emph{human-centered} perspective to provide guidance for robot design so that human teachers can naturally teach ITL robots. In this paper, we present 1) a qualitative bidirectional analysis of an interactive teaching study (N=10) through which we characterize various aspects of actions intended and executed by human teachers when teaching a robot; 2) an in-depth discussion of the teaching approach employed by two participants to understand the need for personal adaptation to individual teaching styles; and 3) requirements for ITL robot design based on our analyses and informed by a computational theory of collaborative interactions, SharedPlans.
Most #AI, #ML research assumes the first configuration in the figure below. We take it for granted that a human will fully delegate a task to the #AI #system
That couldn't be further from reality. For a long time, humans and #AI/#ML systems will have to work together in various configurations.
#CollaborativeHumanAISystems looks at what is typically overlooked in #AI, #ML - how do we bring humans in the #AI loop.
Shiwali Mohan | शिवाली मोहन