Research Team Leaders and Projects
Research Project 1: Hoda Heidari (CS)
Hoda Heidari, Assistant Professor, Machine Learning and the Institute for Software, Systems, and Society & Anna Kawakami, PhD Assistant
We will develop an automated web crawling tool that scrapes federal and state-level government websites to identify (and later survey) regulations concerning the use of Artificial Intelligence and data-driven algorithms.
We will develop an automated web crawling tool that scrapes federal and state-level government websites to identify (and later survey) regulations concerning the use of Artificial Intelligence and data-driven algorithms.
Research Project 2: Srinivasan Seshan (CS)
Srinivasan Seshan, Joseph F. Traub Professor and Department Head, Computer Science
How fair is the internet? The design of the Internet relies on a very simple processing inside the network and more complex tasks being performed by its endpoints. One task that is surprisingly left to endpoints to implement is resource and bandwidth allocation. End points run congestion control algorithms (CCAs) that make observations about the network (e.g. delay and packet loss) to decide how fast to transmit data on any connection. These algorithms and their correct operation are critical to both the stability of the Internet as well as how fairly bandwidth resources are divided amongst its users. During most of the history of the Internet, there have been just a few CCAs widely deployed and their behavior and interactions were well understood. However, over the past decade there has been growing interest in CCA design and as a result many new algorithms have been developed and widely deployed (e.g. BBR by Google). Unfortunately, our understanding of how all these algorithms compete with each other and the resulting fairness of resource allocation is surprisingly poor. In this project, participants will learn the basics of congestion control and how to perform network monitoring. They will then design and evaluate a tool that evaluates the fairness of both pre-recorded network traces and live traffic. Participants will only require basic programming skills in Python — no networking experience needed.
How fair is the internet? The design of the Internet relies on a very simple processing inside the network and more complex tasks being performed by its endpoints. One task that is surprisingly left to endpoints to implement is resource and bandwidth allocation. End points run congestion control algorithms (CCAs) that make observations about the network (e.g. delay and packet loss) to decide how fast to transmit data on any connection. These algorithms and their correct operation are critical to both the stability of the Internet as well as how fairly bandwidth resources are divided amongst its users. During most of the history of the Internet, there have been just a few CCAs widely deployed and their behavior and interactions were well understood. However, over the past decade there has been growing interest in CCA design and as a result many new algorithms have been developed and widely deployed (e.g. BBR by Google). Unfortunately, our understanding of how all these algorithms compete with each other and the resulting fairness of resource allocation is surprisingly poor. In this project, participants will learn the basics of congestion control and how to perform network monitoring. They will then design and evaluate a tool that evaluates the fairness of both pre-recorded network traces and live traffic. Participants will only require basic programming skills in Python — no networking experience needed.
Research Project 3: Lorrie Faith Cranor (CS & EPP)
Lorrie Faith Cranor, Director and Bosch Distinguished Professor in Security and Privacy Technologies, CyLab Security and Privacy Institute, FORE Systems Professor, Computer Science and Engineering & Public Policy, Elijah Bouma-Sims, PhD Assistant and Ally Nisenoff, PhD Assistant
Cookie consent interfaces have become ubiquitous on the internet as websites attempt to comply with global privacy regulations. Some of these interfaces allow users to opt in to or opt out of allowing cookies that are used for specific categories of purposes (for example, functional cookies, performance cookies, strictly necessary cookies, targeting cookies). These cookie categories were introduced over 10 years ago and have not changed since. However, research has found that one of the barriers to user success in decision making about cookies is a lack of understanding about the meanings of cookie terms. In this project we will discuss alternative names for cookie categories, and then design and carry out an online survey using a crowdworker platform to determine whether any of our alternative category names improve user understanding about cookies.
Cookie consent interfaces have become ubiquitous on the internet as websites attempt to comply with global privacy regulations. Some of these interfaces allow users to opt in to or opt out of allowing cookies that are used for specific categories of purposes (for example, functional cookies, performance cookies, strictly necessary cookies, targeting cookies). These cookie categories were introduced over 10 years ago and have not changed since. However, research has found that one of the barriers to user success in decision making about cookies is a lack of understanding about the meanings of cookie terms. In this project we will discuss alternative names for cookie categories, and then design and carry out an online survey using a crowdworker platform to determine whether any of our alternative category names improve user understanding about cookies.
Research Project 4: Mallesham Dasari (CS) & Anthony Rowe (ECE)
Mallesham Dasari, Postdoc, Computer Science and Anthony Rowe, Siewiorek and Walker Family Professor, Electrical and Computer Engineering and CyLab
Capturing and streaming 3D space. It has long been a goal of immersive telepresence to capture and stream 3D spaces such that a remote viewer can watch from any location or angle within the scene. In this project, we will use off-the-shelf software and hardware components such as 3D cameras with a set of graphics pipelines to capture 3D scenes and stream over a network with bandwidth constraints. Over the course of the two day workshop, we will understand the requirements of streaming 3D scenes and tackle a set of practical challenges (such as latency and bandwidth) involved in streaming 3D scenes. If time permits, we will devise potential optimizations to address these challenges and implement one or two to show the performance benefits of the proposed optimizations.
Capturing and streaming 3D space. It has long been a goal of immersive telepresence to capture and stream 3D spaces such that a remote viewer can watch from any location or angle within the scene. In this project, we will use off-the-shelf software and hardware components such as 3D cameras with a set of graphics pipelines to capture 3D scenes and stream over a network with bandwidth constraints. Over the course of the two day workshop, we will understand the requirements of streaming 3D scenes and tackle a set of practical challenges (such as latency and bandwidth) involved in streaming 3D scenes. If time permits, we will devise potential optimizations to address these challenges and implement one or two to show the performance benefits of the proposed optimizations.
Research Project 5: Aruna Amin & Jing Huang (Walmart Global Tech)
Aruna Amin, Director, Data Engineering & Jing Huang, Director of Applied Science, Transportation and Last Mile (Walmart Global Tech)
Optimizing driver dispatching for crowd sourced delivery platform. The crowdsourced delivery platform for last mile delivery provides contracted drivers with a flexible way of earning money by shopping and delivering customer orders from retailers. Here is how it works: when delivery time window approaches, the system will publish trips to registered drivers. When receiving trips, drivers can make the decision on whether to accept them. Once drivers choose to accept the trip, they will drive to the pickup location, pick up orders and make deliveries. Driver dispatching is to make real-time decision on how to match drivers effectively and efficiently with trips. It is crucial to the health of last mile eco-system by optimizing drivers’ efficiency to unlock their earning potential while providing a consistently satisfying customer experience. This research problem will be focusing on using various ML and optimization models to efficiently solve for driver trip matching. And example questions to explore include: how to understand driver’s engagement and affinity to certain types of trips? How to maximize the success rate when matching a driver to a trip? How to deal with trade-offs among multiple objectives such as matching success rate, trip efficiency and drivers' engagement?
Optimizing driver dispatching for crowd sourced delivery platform. The crowdsourced delivery platform for last mile delivery provides contracted drivers with a flexible way of earning money by shopping and delivering customer orders from retailers. Here is how it works: when delivery time window approaches, the system will publish trips to registered drivers. When receiving trips, drivers can make the decision on whether to accept them. Once drivers choose to accept the trip, they will drive to the pickup location, pick up orders and make deliveries. Driver dispatching is to make real-time decision on how to match drivers effectively and efficiently with trips. It is crucial to the health of last mile eco-system by optimizing drivers’ efficiency to unlock their earning potential while providing a consistently satisfying customer experience. This research problem will be focusing on using various ML and optimization models to efficiently solve for driver trip matching. And example questions to explore include: how to understand driver’s engagement and affinity to certain types of trips? How to maximize the success rate when matching a driver to a trip? How to deal with trade-offs among multiple objectives such as matching success rate, trip efficiency and drivers' engagement?
Research Project 6: Aruna Amin & Manali Walke (Walmart Global Tech)
Aruna Amin, Director, Data Engineering & Manali Walke, Data Scientist at Walmart Labs (Walmart Global Tech)
'Using ML to automate store operations. Machine learning has been an important and fiercely growing aspect in nearly all fields. Stores, however, don't seem to use as much artificial intelligence on the surface, with most of them operating in a regular supermarket fashion. Stores are a treasure trove for machine learning and artificial applications, right from automating inventory accuracy, detecting stale or spoiled goods and maybe even disposing/recycling them! This research problem is for exploring not only such use cases but also caveats and privacy concerns regarding them.
'Using ML to automate store operations. Machine learning has been an important and fiercely growing aspect in nearly all fields. Stores, however, don't seem to use as much artificial intelligence on the surface, with most of them operating in a regular supermarket fashion. Stores are a treasure trove for machine learning and artificial applications, right from automating inventory accuracy, detecting stale or spoiled goods and maybe even disposing/recycling them! This research problem is for exploring not only such use cases but also caveats and privacy concerns regarding them.
Research Project 7: Anh Truong (Adobe Research)
Anh Truong, Creative Intelligence Lab at Adobe Research
Trend based music montage videos have become extremely popular on short-form video sharing platforms such as TikTok, Instagram Reels, and YouTube shorts. These videos generally consist of an audio segment edited with multiple clips of visual media that are cut together in a specific way to tell a short story. When such a video goes viral, other creators with try to replicate the original video by using the same audio and general structure with different visuals to tell a similar, but personalized story. However, this replication process is often tedious and time consuming. Although there is structure from the exemplar video that these new creators want to emulate, existing tools for editing short form videos require the user to create the new video from scratch and without the context of the exemplar video. In this project, we will conduct a formative study of popular short-form music montage videos and the tools used to create them. Our goal is to identify the common structure between the exemplars and replica videos, as well as interaction bottlenecks of existing tools. Using these findings, we will design and implement a template-based tool that exploits the structure of an exemplar video to make this trendy music montage video creation process easier.
Trend based music montage videos have become extremely popular on short-form video sharing platforms such as TikTok, Instagram Reels, and YouTube shorts. These videos generally consist of an audio segment edited with multiple clips of visual media that are cut together in a specific way to tell a short story. When such a video goes viral, other creators with try to replicate the original video by using the same audio and general structure with different visuals to tell a similar, but personalized story. However, this replication process is often tedious and time consuming. Although there is structure from the exemplar video that these new creators want to emulate, existing tools for editing short form videos require the user to create the new video from scratch and without the context of the exemplar video. In this project, we will conduct a formative study of popular short-form music montage videos and the tools used to create them. Our goal is to identify the common structure between the exemplars and replica videos, as well as interaction bottlenecks of existing tools. Using these findings, we will design and implement a template-based tool that exploits the structure of an exemplar video to make this trendy music montage video creation process easier.
Research Project 8: Nur Yildirim (HCII)
Nur Yildirim, Doctoral Student, Human-Computer Interaction
Brainstorming Human-AI interactions. Advances in Artificial intelligence (AI) have enabled amazing technology capabilities: computers can diagnose diseases, translate between languages, and drive cars. However, most AI systems fail to produce value for humans, and even cause unintended harm. This workshop explores brainstorming with AI capabilities to identify opportunities where AI can improve our daily lives, and to anticipate how it can lead to harm. The workshop will combine a short introductory lecture on AI capabilities and brainstorming sessions for generating and detailing AI concepts. At the end of the workshop, the team will have generated an AI concept to address a problem in daily life.
Research Project 9: Hedyeh Beyhaghi (CS)
Hedyeh Beyhaghi, Postdoc, Machine Learning
Exploring problems in algorithmic game theory. In recent years, game theory, a field originally developed in economics, has substantially impacted computer science. This impact has resulted in an interdisciplinary field called algorithmic game theory, which studies algorithms in strategic environments. In this workshop, we learn about some of the foundational problems in algorithmic game theory, such as selfish routing and the secretary problem. We use these problems as case studies and practice different phases of theoretical research, from mathematical modeling to developing and evaluating conjectures and proofs.
Exploring problems in algorithmic game theory. In recent years, game theory, a field originally developed in economics, has substantially impacted computer science. This impact has resulted in an interdisciplinary field called algorithmic game theory, which studies algorithms in strategic environments. In this workshop, we learn about some of the foundational problems in algorithmic game theory, such as selfish routing and the secretary problem. We use these problems as case studies and practice different phases of theoretical research, from mathematical modeling to developing and evaluating conjectures and proofs.
Research Project 10: Hope Chidziwisano (HCII)
Hope Chidziwisano, President's Postdoctoral Fellow, Human Computer Interaction Institute
Smart Home Security Systems and Women’s Privacy in Patriarchal Societies. According to the World Bank, one in three women in patriarchal societies experience domestic violence due to men’s social status over them. Prior research suggests that the introduction of smart home security systems in these homes can perpetuate patriarchal attitudes and exacerbate the privacy challenges women face thereby worsening domestic violence against them. Despite this, women’s perspectives on smart home security systems are not well understood. In this workshop, I will provide an overview of the problem, then present an approach I used to understand personal privacy challenges women face when using these technologies in patriarchal societies. Next, I will highlight key opportunities for designing personal privacy-aware home security technologies for patriarchal societies. Based on these design opportunities, we will focus on prototyping privacy-aware smart home security systems for patriarchal societies.
Smart Home Security Systems and Women’s Privacy in Patriarchal Societies. According to the World Bank, one in three women in patriarchal societies experience domestic violence due to men’s social status over them. Prior research suggests that the introduction of smart home security systems in these homes can perpetuate patriarchal attitudes and exacerbate the privacy challenges women face thereby worsening domestic violence against them. Despite this, women’s perspectives on smart home security systems are not well understood. In this workshop, I will provide an overview of the problem, then present an approach I used to understand personal privacy challenges women face when using these technologies in patriarchal societies. Next, I will highlight key opportunities for designing personal privacy-aware home security technologies for patriarchal societies. Based on these design opportunities, we will focus on prototyping privacy-aware smart home security systems for patriarchal societies.