Human-computer interaction

Subterms

    Latest story

    138 Shares119 Views

    Method prevents an AI model from being overconfident about wrong answers

    by

    People use large language models for a huge array of tasks, from translating an article to identifying financial fraud. However, despite the incredible capabilities and versatility of these models, they sometimes generate inaccurate responses.On top of that problem, the models can be overconfident about wrong answers or underconfident about correct ones, making it tough for a user to know when a model can be trusted.Researchers typically calibrate a machine-learning model to ensure its level of confidence lines up with its accuracy. A well-calibrated model should have less confidence about an incorrect prediction, and vice-versa. But because large language models (LLMs) can be applied to a seemingly endless collection of diverse tasks, traditional calibration methods are ineffective.Now, researchers from MIT and the MIT-IBM Watson AI Lab have introduced a calibration method tailored to large language models. Their method, called Thermometer, involves building a smaller, auxiliary model that runs on top of a large language model to calibrate it.Thermometer is more efficient than other approaches — requiring less power-hungry computation — while preserving the accuracy of the model and enabling it to produce better-calibrated responses on tasks it has not seen before.By enabling efficient calibration of an LLM for a variety of tasks, Thermometer could help users pinpoint situations where a model is overconfident about false predictions, ultimately preventing them from deploying that model in a situation where it may fail.“With Thermometer, we want to provide the user with a clear signal to tell them whether a model’s response is accurate or inaccurate, in a way that reflects the model’s uncertainty, so they know if that model is reliable,” says Maohao Shen, an electrical engineering and computer science (EECS) graduate student and lead author of a paper on Thermometer.Shen is joined on the paper by Gregory Wornell, the Sumitomo Professor of Engineering who leads the Signals, Information, and Algorithms Laboratory in the Research Laboratory for Electronics, and is a member of the MIT-IBM Watson AI Lab; senior author Soumya Ghosh, a research staff member in the MIT-IBM Watson AI Lab; as well as others at MIT and the MIT-IBM Watson AI Lab. The research was recently presented at the International Conference on Machine Learning.Universal calibrationSince traditional machine-learning models are typically designed to perform a single task, calibrating them usually involves one task-specific method. On the other hand, since LLMs have the flexibility to perform many tasks, using a traditional method to calibrate that model for one task might hurt its performance on another task.Calibrating an LLM often involves sampling from the model multiple times to obtain different predictions and then aggregating these predictions to obtain better-calibrated confidence. However, because these models have billions of parameters, the computational costs of such approaches rapidly add up.“In a sense, large language models are universal because they can handle various tasks. So, we need a universal calibration method that can also handle many different tasks,” says Shen.With Thermometer, the researchers developed a versatile technique that leverages a classical calibration method called temperature scaling to efficiently calibrate an LLM for a new task.In this context, a “temperature” is a scaling parameter used to adjust a model’s confidence to be aligned with its prediction accuracy. Traditionally, one determines the right temperature using a labeled validation dataset of task-specific examples.Since LLMs are often applied to new tasks, labeled datasets can be nearly impossible to acquire. For instance, a user who wants to deploy an LLM to answer customer questions about a new product likely does not have a dataset containing such questions and answers.Instead of using a labeled dataset, the researchers train an auxiliary model that runs on top of an LLM to automatically predict the temperature needed to calibrate it for this new task.They use labeled datasets of a few representative tasks to train the Thermometer model, but then once it has been trained, it can generalize to new tasks in a similar category without the need for additional labeled data.A Thermometer model trained on a collection of multiple-choice question datasets, perhaps including one with algebra questions and one with medical questions, could be used to calibrate an LLM that will answer questions about geometry or biology, for instance.“The aspirational goal is for it to work on any task, but we are not quite there yet,” Ghosh says.   The Thermometer model only needs to access a small part of the LLM’s inner workings to predict the right temperature that will calibrate its prediction for data points of a specific task. An efficient approachImportantly, the technique does not require multiple training runs and only slightly slows the LLM. Plus, since temperature scaling does not alter a model’s predictions, Thermometer preserves its accuracy.When they compared Thermometer to several baselines on multiple tasks, it consistently produced better-calibrated uncertainty measures while requiring much less computation.“As long as we train a Thermometer model on a sufficiently large number of tasks, it should be able to generalize well across any new task, just like a large language model, it is also a universal model,” Shen adds.The researchers also found that if they train a Thermometer model for a smaller LLM, it can be directly applied to calibrate a larger LLM within the same family.In the future, they want to adapt Thermometer for more complex text-generation tasks and apply the technique to even larger LLMs. The researchers also hope to quantify the diversity and number of labeled datasets one would need to train a Thermometer model so it can generalize to a new task.This research was funded, in part, by the MIT-IBM Watson AI Lab. More

    More stories

    • 138 Shares189 Views

      in Data Management & Statistics

      Study: When allocating scarce resources with AI, randomization can improve fairness

      by

      Organizations are increasingly utilizing machine-learning models to allocate scarce resources or opportunities. For instance, such models can help companies screen resumes to choose job interview candidates or aid hospitals in ranking kidney transplant patients based on their likelihood of survival.When deploying a model, users typically strive to ensure its predictions are fair by reducing bias. This often involves techniques like adjusting the features a model uses to make decisions or calibrating the scores it generates.However, researchers from MIT and Northeastern University argue that these fairness methods are not sufficient to address structural injustices and inherent uncertainties. In a new paper, they show how randomizing a model’s decisions in a structured way can improve fairness in certain situations.For example, if multiple companies use the same machine-learning model to rank job interview candidates deterministically — without any randomization — then one deserving individual could be the bottom-ranked candidate for every job, perhaps due to how the model weighs answers provided in an online form. Introducing randomization into a model’s decisions could prevent one worthy person or group from always being denied a scarce resource, like a job interview.Through their analysis, the researchers found that randomization can be especially beneficial when a model’s decisions involve uncertainty or when the same group consistently receives negative decisions.They present a framework one could use to introduce a specific amount of randomization into a model’s decisions by allocating resources through a weighted lottery. This method, which an individual can tailor to fit their situation, can improve fairness without hurting the efficiency or accuracy of a model.“Even if you could make fair predictions, should you be deciding these social allocations of scarce resources or opportunities strictly off scores or rankings? As things scale, and we see more and more opportunities being decided by these algorithms, the inherent uncertainties in these scores can be amplified. We show that fairness may require some sort of randomization,” says Shomik Jain, a graduate student in the Institute for Data, Systems, and Society (IDSS) and lead author of the paper.Jain is joined on the paper by Kathleen Creel, assistant professor of philosophy and computer science at Northeastern University; and senior author Ashia Wilson, the Lister Brothers Career Development Professor in the Department of Electrical Engineering and Computer Science and a principal investigator in the Laboratory for Information and Decision Systems (LIDS). The research will be presented at the International Conference on Machine Learning.Considering claimsThis work builds off a previous paper in which the researchers explored harms that can occur when one uses deterministic systems at scale. They found that using a machine-learning model to deterministically allocate resources can amplify inequalities that exist in training data, which can reinforce bias and systemic inequality. “Randomization is a very useful concept in statistics, and to our delight, satisfies the fairness demands coming from both a systemic and individual point of view,” Wilson says.In this paper, they explored the question of when randomization can improve fairness. They framed their analysis around the ideas of philosopher John Broome, who wrote about the value of using lotteries to award scarce resources in a way that honors all claims of individuals.A person’s claim to a scarce resource, like a kidney transplant, can stem from merit, deservingness, or need. For instance, everyone has a right to life, and their claims on a kidney transplant may stem from that right, Wilson explains.“When you acknowledge that people have different claims to these scarce resources, fairness is going to require that we respect all claims of individuals. If we always give someone with a stronger claim the resource, is that fair?” Jain says.That sort of deterministic allocation could cause systemic exclusion or exacerbate patterned inequality, which occurs when receiving one allocation increases an individual’s likelihood of receiving future allocations. In addition, machine-learning models can make mistakes, and a deterministic approach could cause the same mistake to be repeated.Randomization can overcome these problems, but that doesn’t mean all decisions a model makes should be randomized equally.Structured randomizationThe researchers use a weighted lottery to adjust the level of randomization based on the amount of uncertainty involved in the model’s decision-making. A decision that is less certain should incorporate more randomization.“In kidney allocation, usually the planning is around projected lifespan, and that is deeply uncertain. If two patients are only five years apart, it becomes a lot harder to measure. We want to leverage that level of uncertainty to tailor the randomization,” Wilson says.The researchers used statistical uncertainty quantification methods to determine how much randomization is needed in different situations. They show that calibrated randomization can lead to fairer outcomes for individuals without significantly affecting the utility, or effectiveness, of the model.“There is a balance to be had between overall utility and respecting the rights of the individuals who are receiving a scarce resource, but oftentimes the tradeoff is relatively small,” says Wilson.However, the researchers emphasize there are situations where randomizing decisions would not improve fairness and could harm individuals, such as in criminal justice contexts.But there could be other areas where randomization can improve fairness, such as college admissions, and the researchers plan to study other use cases in future work. They also want to explore how randomization can affect other factors, such as competition or prices, and how it could be used to improve the robustness of machine-learning models.“We are hoping our paper is a first move toward illustrating that there might be a benefit to randomization. We are offering randomization as a tool. How much you are going to want to do it is going to be up to all the stakeholders in the allocation to decide. And, of course, how they decide is another research question all together,” says Wilson. More

    • 138 Shares109 Views

      in Data Management & Statistics

      How to assess a general-purpose AI model’s reliability before it’s deployed

      by

      Foundation models are massive deep-learning models that have been pretrained on an enormous amount of general-purpose, unlabeled data. They can be applied to a variety of tasks, like generating images or answering customer questions.But these models, which serve as the backbone for powerful artificial intelligence tools like ChatGPT and DALL-E, can offer up incorrect or misleading information. In a safety-critical situation, such as a pedestrian approaching a self-driving car, these mistakes could have serious consequences.To help prevent such mistakes, researchers from MIT and the MIT-IBM Watson AI Lab developed a technique to estimate the reliability of foundation models before they are deployed to a specific task.They do this by considering a set of foundation models that are slightly different from one another. Then they use their algorithm to assess the consistency of the representations each model learns about the same test data point. If the representations are consistent, it means the model is reliable.When they compared their technique to state-of-the-art baseline methods, it was better at capturing the reliability of foundation models on a variety of downstream classification tasks.Someone could use this technique to decide if a model should be applied in a certain setting, without the need to test it on a real-world dataset. This could be especially useful when datasets may not be accessible due to privacy concerns, like in health care settings. In addition, the technique could be used to rank models based on reliability scores, enabling a user to select the best one for their task.“All models can be wrong, but models that know when they are wrong are more useful. The problem of quantifying uncertainty or reliability is more challenging for these foundation models because their abstract representations are difficult to compare. Our method allows one to quantify how reliable a representation model is for any given input data,” says senior author Navid Azizan, the Esther and Harold E. Edgerton Assistant Professor in the MIT Department of Mechanical Engineering and the Institute for Data, Systems, and Society (IDSS), and a member of the Laboratory for Information and Decision Systems (LIDS).He is joined on a paper about the work by lead author Young-Jin Park, a LIDS graduate student; Hao Wang, a research scientist at the MIT-IBM Watson AI Lab; and Shervin Ardeshir, a senior research scientist at Netflix. The paper will be presented at the Conference on Uncertainty in Artificial Intelligence.Measuring consensusTraditional machine-learning models are trained to perform a specific task. These models typically make a concrete prediction based on an input. For instance, the model might tell you whether a certain image contains a cat or a dog. In this case, assessing reliability could be a matter of looking at the final prediction to see if the model is right.But foundation models are different. The model is pretrained using general data, in a setting where its creators don’t know all downstream tasks it will be applied to. Users adapt it to their specific tasks after it has already been trained.Unlike traditional machine-learning models, foundation models don’t give concrete outputs like “cat” or “dog” labels. Instead, they generate an abstract representation based on an input data point.To assess the reliability of a foundation model, the researchers used an ensemble approach by training several models which share many properties but are slightly different from one another.“Our idea is like measuring the consensus. If all those foundation models are giving consistent representations for any data in our dataset, then we can say this model is reliable,” Park says.But they ran into a problem: How could they compare abstract representations?“These models just output a vector, comprised of some numbers, so we can’t compare them easily,” he adds.They solved this problem using an idea called neighborhood consistency.For their approach, the researchers prepare a set of reliable reference points to test on the ensemble of models. Then, for each model, they investigate the reference points located near that model’s representation of the test point.By looking at the consistency of neighboring points, they can estimate the reliability of the models.Aligning the representationsFoundation models map data points to what is known as a representation space. One way to think about this space is as a sphere. Each model maps similar data points to the same part of its sphere, so images of cats go in one place and images of dogs go in another.But each model would map animals differently in its own sphere, so while cats may be grouped near the South Pole of one sphere, another model could map cats somewhere in the Northern Hemisphere.The researchers use the neighboring points like anchors to align those spheres so they can make the representations comparable. If a data point’s neighbors are consistent across multiple representations, then one should be confident about the reliability of the model’s output for that point.When they tested this approach on a wide range of classification tasks, they found that it was much more consistent than baselines. Plus, it wasn’t tripped up by challenging test points that caused other methods to fail.Moreover, their approach can be used to assess reliability for any input data, so one could evaluate how well a model works for a particular type of individual, such as a patient with certain characteristics.“Even if the models all have average performance overall, from an individual point of view, you’d prefer the one that works best for that individual,” Wang says.However, one limitation comes from the fact that they must train an ensemble of foundation models, which is computationally expensive. In the future, they plan to find more efficient ways to build multiple models, perhaps by using small perturbations of a single model.“With the current trend of using foundational models for their embeddings to support various downstream tasks — from fine-tuning to retrieval augmented generation — the topic of quantifying uncertainty at the representation level is increasingly important, but challenging, as embeddings on their own have no grounding. What matters instead is how embeddings of different inputs are related to one another, an idea that this work neatly captures through the proposed neighborhood consistency score,” says Marco Pavone, an associate professor in the Department of Aeronautics and Astronautics at Stanford University, who was not involved with this work. “This is a promising step towards high quality uncertainty quantifications for embedding models, and I’m excited to see future extensions which can operate without requiring model-ensembling to really enable this approach to scale to foundation-size models.”This work is funded, in part, by the MIT-IBM Watson AI Lab, MathWorks, and Amazon. More

    • 75 Shares169 Views

      in Data Management & Statistics

      When to trust an AI model

      by

      Because machine-learning models can give false predictions, researchers often equip them with the ability to tell a user how confident they are about a certain decision. This is especially important in high-stake settings, such as when models are used to help identify disease in medical images or filter job applications.But a model’s uncertainty quantifications are only useful if they are accurate. If a model says it is 49 percent confident that a medical image shows a pleural effusion, then 49 percent of the time, the model should be right.MIT researchers have introduced a new approach that can improve uncertainty estimates in machine-learning models. Their method not only generates more accurate uncertainty estimates than other techniques, but does so more efficiently.In addition, because the technique is scalable, it can be applied to huge deep-learning models that are increasingly being deployed in health care and other safety-critical situations.This technique could give end users, many of whom lack machine-learning expertise, better information they can use to determine whether to trust a model’s predictions or if the model should be deployed for a particular task.“It is easy to see these models perform really well in scenarios where they are very good, and then assume they will be just as good in other scenarios. This makes it especially important to push this kind of work that seeks to better calibrate the uncertainty of these models to make sure they align with human notions of uncertainty,” says lead author Nathan Ng, a graduate student at the University of Toronto who is a visiting student at MIT.Ng wrote the paper with Roger Grosse, an assistant professor of computer science at the University of Toronto; and senior author Marzyeh Ghassemi, an associate professor in the Department of Electrical Engineering and Computer Science and a member of the Institute of Medical Engineering Sciences and the Laboratory for Information and Decision Systems. The research will be presented at the International Conference on Machine Learning.Quantifying uncertaintyUncertainty quantification methods often require complex statistical calculations that don’t scale well to machine-learning models with millions of parameters. These methods also require users to make assumptions about the model and data used to train it.The MIT researchers took a different approach. They use what is known as the minimum description length principle (MDL), which does not require the assumptions that can hamper the accuracy of other methods. MDL is used to better quantify and calibrate uncertainty for test points the model has been asked to label.The technique the researchers developed, known as IF-COMP, makes MDL fast enough to use with the kinds of large deep-learning models deployed in many real-world settings.MDL involves considering all possible labels a model could give a test point. If there are many alternative labels for this point that fit well, its confidence in the label it chose should decrease accordingly.“One way to understand how confident a model is would be to tell it some counterfactual information and see how likely it is to believe you,” Ng says.For example, consider a model that says a medical image shows a pleural effusion. If the researchers tell the model this image shows an edema, and it is willing to update its belief, then the model should be less confident in its original decision.With MDL, if a model is confident when it labels a datapoint, it should use a very short code to describe that point. If it is uncertain about its decision because the point could have many other labels, it uses a longer code to capture these possibilities.The amount of code used to label a datapoint is known as stochastic data complexity. If the researchers ask the model how willing it is to update its belief about a datapoint given contrary evidence, the stochastic data complexity should decrease if the model is confident.But testing each datapoint using MDL would require an enormous amount of computation.Speeding up the processWith IF-COMP, the researchers developed an approximation technique that can accurately estimate stochastic data complexity using a special function, known as an influence function. They also employed a statistical technique called temperature-scaling, which improves the calibration of the model’s outputs. This combination of influence functions and temperature-scaling enables high-quality approximations of the stochastic data complexity.In the end, IF-COMP can efficiently produce well-calibrated uncertainty quantifications that reflect a model’s true confidence. The technique can also determine whether the model has mislabeled certain data points or reveal which data points are outliers.The researchers tested their system on these three tasks and found that it was faster and more accurate than other methods.“It is really important to have some certainty that a model is well-calibrated, and there is a growing need to detect when a specific prediction doesn’t look quite right. Auditing tools are becoming more necessary in machine-learning problems as we use large amounts of unexamined data to make models that will be applied to human-facing problems,” Ghassemi says.IF-COMP is model-agnostic, so it can provide accurate uncertainty quantifications for many types of machine-learning models. This could enable it to be deployed in a wider range of real-world settings, ultimately helping more practitioners make better decisions.“People need to understand that these systems are very fallible and can make things up as they go. A model may look like it is highly confident, but there are a ton of different things it is willing to believe given evidence to the contrary,” Ng says.In the future, the researchers are interested in applying their approach to large language models and studying other potential use cases for the minimum description length principle.  More

    • 125 Shares179 Views

      in Data Management & Statistics

      “They can see themselves shaping the world they live in”

      by

      During the journey from the suburbs to the city, the tree canopy often dwindles down as skyscrapers rise up. A group of New England Innovation Academy students wondered why that is.“Our friend Victoria noticed that where we live in Marlborough there are lots of trees in our own backyards. But if you drive just 30 minutes to Boston, there are almost no trees,” said high school junior Ileana Fournier. “We were struck by that duality.”This inspired Fournier and her classmates Victoria Leeth and Jessie Magenyi to prototype a mobile app that illustrates Massachusetts deforestation trends for Day of AI, a free, hands-on curriculum developed by the MIT Responsible AI for Social Empowerment and Education (RAISE) initiative, headquartered in the MIT Media Lab and in collaboration with the MIT Schwarzman College of Computing and MIT Open Learning. They were among a group of 20 students from New England Innovation Academy who shared their projects during the 2024 Day of AI global celebration hosted with the Museum of Science.The Day of AI curriculum introduces K-12 students to artificial intelligence. Now in its third year, Day of AI enables students to improve their communities and collaborate on larger global challenges using AI. Fournier, Leeth, and Magenyi’s TreeSavers app falls under the Telling Climate Stories with Data module, one of four new climate-change-focused lessons.“We want you to be able to express yourselves creatively to use AI to solve problems with critical-thinking skills,” Cynthia Breazeal, director of MIT RAISE, dean for digital learning at MIT Open Learning, and professor of media arts and sciences, said during this year’s Day of AI global celebration at the Museum of Science. “We want you to have an ethical and responsible way to think about this really powerful, cool, and exciting technology.”Moving from understanding to actionDay of AI invites students to examine the intersection of AI and various disciplines, such as history, civics, computer science, math, and climate change. With the curriculum available year-round, more than 10,000 educators across 114 countries have brought Day of AI activities to their classrooms and homes.The curriculum gives students the agency to evaluate local issues and invent meaningful solutions. “We’re thinking about how to create tools that will allow kids to have direct access to data and have a personal connection that intersects with their lived experiences,” Robert Parks, curriculum developer at MIT RAISE, said at the Day of AI global celebration.Before this year, first-year Jeremie Kwapong said he knew very little about AI. “I was very intrigued,” he said. “I started to experiment with ChatGPT to see how it reacts. How close can I get this to human emotion? What is AI’s knowledge compared to a human’s knowledge?”In addition to helping students spark an interest in AI literacy, teachers around the world have told MIT RAISE that they want to use data science lessons to engage students in conversations about climate change. Therefore, Day of AI’s new hands-on projects use weather and climate change to show students why it’s important to develop a critical understanding of dataset design and collection when observing the world around them.“There is a lag between cause and effect in everyday lives,” said Parks. “Our goal is to demystify that, and allow kids to access data so they can see a long view of things.”Tools like MIT App Inventor — which allows anyone to create a mobile application — help students make sense of what they can learn from data. Fournier, Leeth, and Magenyi programmed TreeSavers in App Inventor to chart regional deforestation rates across Massachusetts, identify ongoing trends through statistical models, and predict environmental impact. The students put that “long view” of climate change into practice when developing TreeSavers’ interactive maps. Users can toggle between Massachusetts’s current tree cover, historical data, and future high-risk areas.Although AI provides fast answers, it doesn’t necessarily offer equitable solutions, said David Sittenfeld, director of the Center for the Environment at the Museum of Science. The Day of AI curriculum asks students to make decisions on sourcing data, ensuring unbiased data, and thinking responsibly about how findings could be used.“There’s an ethical concern about tracking people’s data,” said Ethan Jorda, a New England Innovation Academy student. His group used open-source data to program an app that helps users track and reduce their carbon footprint.Christine Cunningham, senior vice president of STEM Learning at the Museum of Science, believes students are prepared to use AI responsibly to make the world a better place. “They can see themselves shaping the world they live in,” said Cunningham. “Moving through from understanding to action, kids will never look at a bridge or a piece of plastic lying on the ground in the same way again.”Deepening collaboration on earth and beyondThe 2024 Day of AI speakers emphasized collaborative problem solving at the local, national, and global levels.“Through different ideas and different perspectives, we’re going to get better solutions,” said Cunningham. “How do we start young enough that every child has a chance to both understand the world around them but also to move toward shaping the future?”Presenters from MIT, the Museum of Science, and NASA approached this question with a common goal — expanding STEM education to learners of all ages and backgrounds.“We have been delighted to collaborate with the MIT RAISE team to bring this year’s Day of AI celebration to the Museum of Science,” says Meg Rosenburg, manager of operations at the Museum of Science Centers for Public Science Learning. “This opportunity to highlight the new climate modules for the curriculum not only perfectly aligns with the museum’s goals to focus on climate and active hope throughout our Year of the Earthshot initiative, but it has also allowed us to bring our teams together and grow a relationship that we are very excited to build upon in the future.”Rachel Connolly, systems integration and analysis lead for NASA’s Science Activation Program, showed the power of collaboration with the example of how human comprehension of Saturn’s appearance has evolved. From Galileo’s early telescope to the Cassini space probe, modern imaging of Saturn represents 400 years of science, technology, and math working together to further knowledge.“Technologies, and the engineers who built them, advance the questions we’re able to ask and therefore what we’re able to understand,” said Connolly, research scientist at MIT Media Lab.New England Innovation Academy students saw an opportunity for collaboration a little closer to home. Emmett Buck-Thompson, Jeff Cheng, and Max Hunt envisioned a social media app to connect volunteers with local charities. Their project was inspired by Buck-Thompson’s father’s difficulties finding volunteering opportunities, Hunt’s role as the president of the school’s Community Impact Club, and Cheng’s aspiration to reduce screen time for social media users. Using MIT App Inventor, ​their combined ideas led to a prototype with the potential to make a real-world impact in their community.The Day of AI curriculum teaches the mechanics of AI, ethical considerations and responsible uses, and interdisciplinary applications for different fields. It also empowers students to become creative problem solvers and engaged citizens in their communities and online. From supporting volunteer efforts to encouraging action for the state’s forests to tackling the global challenge of climate change, today’s students are becoming tomorrow’s leaders with Day of AI.“We want to empower you to know that this is a tool you can use to make your community better, to help people around you with this technology,” said Breazeal.Other Day of AI speakers included Tim Ritchie, president of the Museum of Science; Michael Lawrence Evans, program director of the Boston Mayor’s Office of New Urban Mechanics; Dava Newman, director of the MIT Media Lab; and Natalie Lao, executive director of the App Inventor Foundation. More

    • 150 Shares189 Views

      in Data Management & Statistics

      New software enables blind and low-vision users to create interactive, accessible charts

      by

      A growing number of tools enable users to make online data representations, like charts, that are accessible for people who are blind or have low vision. However, most tools require an existing visual chart that can then be converted into an accessible format.

      This creates barriers that prevent blind and low-vision users from building their own custom data representations, and it can limit their ability to explore and analyze important information.

      A team of researchers from MIT and University College London (UCL) wants to change the way people think about accessible data representations.

      They created a software system called Umwelt (which means “environment” in German) that can enable blind and low-vision users to build customized, multimodal data representations without needing an initial visual chart.

      Umwelt, an authoring environment designed for screen-reader users, incorporates an editor that allows someone to upload a dataset and create a customized representation, such as a scatterplot, that can include three modalities: visualization, textual description, and sonification. Sonification involves converting data into nonspeech audio.

      The system, which can represent a variety of data types, includes a viewer that enables a blind or low-vision user to interactively explore a data representation, seamlessly switching between each modality to interact with data in a different way.

      The researchers conducted a study with five expert screen-reader users who found Umwelt to be useful and easy to learn. In addition to offering an interface that empowered them to create data representations — something they said was sorely lacking — the users said Umwelt could facilitate communication between people who rely on different senses.

      “We have to remember that blind and low-vision people aren’t isolated. They exist in these contexts where they want to talk to other people about data,” says Jonathan Zong, an electrical engineering and computer science (EECS) graduate student and lead author of a paper introducing Umwelt. “I am hopeful that Umwelt helps shift the way that researchers think about accessible data analysis. Enabling the full participation of blind and low-vision people in data analysis involves seeing visualization as just one piece of this bigger, multisensory puzzle.”

      Joining Zong on the paper are fellow EECS graduate students Isabella Pedraza Pineros and Mengzhu “Katie” Chen; Daniel Hajas, a UCL researcher who works with the Global Disability Innovation Hub; and senior author Arvind Satyanarayan, associate professor of computer science at MIT who leads the Visualization Group in the Computer Science and Artificial Intelligence Laboratory. The paper will be presented at the ACM Conference on Human Factors in Computing.

      De-centering visualization

      The researchers previously developed interactive interfaces that provide a richer experience for screen reader users as they explore accessible data representations. Through that work, they realized most tools for creating such representations involve converting existing visual charts.

      Aiming to decenter visual representations in data analysis, Zong and Hajas, who lost his sight at age 16, began co-designing Umwelt more than a year ago.

      At the outset, they realized they would need to rethink how to represent the same data using visual, auditory, and textual forms.

      “We had to put a common denominator behind the three modalities. By creating this new language for representations, and making the output and input accessible, the whole is greater than the sum of its parts,” says Hajas.

      To build Umwelt, they first considered what is unique about the way people use each sense.

      For instance, a sighted user can see the overall pattern of a scatterplot and, at the same time, move their eyes to focus on different data points. But for someone listening to a sonification, the experience is linear since data are converted into tones that must be played back one at a time.

      “If you are only thinking about directly translating visual features into nonvisual features, then you miss out on the unique strengths and weaknesses of each modality,” Zong adds.

      They designed Umwelt to offer flexibility, enabling a user to switch between modalities easily when one would better suit their task at a given time.

      To use the editor, one uploads a dataset to Umwelt, which employs heuristics to automatically creates default representations in each modality.

      If the dataset contains stock prices for companies, Umwelt might generate a multiseries line chart, a textual structure that groups data by ticker symbol and date, and a sonification that uses tone length to represent the price for each date, arranged by ticker symbol.

      The default heuristics are intended to help the user get started.

      “In any kind of creative tool, you have a blank-slate effect where it is hard to know how to begin. That is compounded in a multimodal tool because you have to specify things in three different representations,” Zong says.

      The editor links interactions across modalities, so if a user changes the textual description, that information is adjusted in the corresponding sonification. Someone could utilize the editor to build a multimodal representation, switch to the viewer for an initial exploration, then return to the editor to make adjustments.

      Helping users communicate about data

      To test Umwelt, they created a diverse set of multimodal representations, from scatterplots to multiview charts, to ensure the system could effectively represent different data types. Then they put the tool in the hands of five expert screen reader users.

      Study participants mostly found Umwelt to be useful for creating, exploring, and discussing data representations. One user said Umwelt was like an “enabler” that decreased the time it took them to analyze data. The users agreed that Umwelt could help them communicate about data more easily with sighted colleagues.

      “What stands out about Umwelt is its core philosophy of de-emphasizing the visual in favor of a balanced, multisensory data experience. Often, nonvisual data representations are relegated to the status of secondary considerations, mere add-ons to their visual counterparts. However, visualization is merely one aspect of data representation. I appreciate their efforts in shifting this perception and embracing a more inclusive approach to data science,” says JooYoung Seo, an assistant professor in the School of Information Sciences at the University of Illinois at Urbana-Champagne, who was not involved with this work.

      Moving forward, the researchers plan to create an open-source version of Umwelt that others can build upon. They also want to integrate tactile sensing into the software system as an additional modality, enabling the use of tools like refreshable tactile graphics displays.

      “In addition to its impact on end users, I am hoping that Umwelt can be a platform for asking scientific questions around how people use and perceive multimodal representations, and how we can improve the design beyond this initial step,” says Zong.

      This work was supported, in part, by the National Science Foundation and the MIT Morningside Academy for Design Fellowship. More

    • 75 Shares159 Views

      in Data Management & Statistics

      “We offer another place for knowledge”

      by

      In the Dzaleka Refugee Camp in Malawi, Jospin Hassan didn’t have access to the education opportunities he sought. So, he decided to create his own. 

      Hassan knew the booming fields of data science and artificial intelligence could bring job opportunities to his community and help solve local challenges. After earning a spot in the 2020-21 cohort of the Certificate Program in Computer and Data Science from MIT Refugee Action Hub (ReACT), Hassan started sharing MIT knowledge and skills with other motivated learners in Dzaleka.

      MIT ReACT is now Emerging Talent, part of the Jameel World Education Lab (J-WEL) at MIT Open Learning. Currently serving its fifth cohort of global learners, Emerging Talent’s year-long certificate program incorporates high-quality computer science and data analysis coursework from MITx, professional skill building, experiential learning, apprenticeship work, and opportunities for networking with MIT’s global community of innovators. Hassan’s cohort honed their leadership skills through interactive online workshops with J-WEL and the 10-week online MIT Innovation Leadership Bootcamp. 

      “My biggest takeaway was networking, collaboration, and learning from each other,” Hassan says.

      Today, Hassan’s organization ADAI Circle offers mentorship and education programs for youth and other job seekers in the Dzaleka Refugee Camp. The curriculum encourages hands-on learning and collaboration.

      Launched in 2020, ADAI Circle aims to foster job creation and reduce poverty in Malawi through technology and innovation. In addition to their classes in data science, AI, software development, and hardware design, their Innovation Hub offers internet access to anyone in need. 

      Doing something different in the community

      Hassan first had the idea for his organization in 2018 when he reached a barrier in his own education journey. There were several programs in the Dzaleka Refugee Camp teaching learners how to code websites and mobile apps, but Hassan felt that they were limited in scope. 

      “We had good devices and internet access,” he says, “but I wanted to learn something new.” 

      Teaming up with co-founder Patrick Byamasu, Hassan and Byamasu set their sights on the longevity of AI and how that might create more jobs for people in their community. “The world is changing every day, and data scientists are in a higher demand today in various companies,” Hassan says. “For this reason, I decided to expand and share the knowledge that I acquired with my fellow refugees and the surrounding villages.”

      ADAI Circle draws inspiration from Hassan’s own experience with MIT Emerging Talent coursework, community, and training opportunities. For example, the MIT Bootcamps model is now standard practice for ADAI Circle’s annual hackathon. Hassan first introduced the hackathon to ADAI Circle students as part of his final experiential learning project of the Emerging Talent certificate program. 

      ADAI Circle’s annual hackathon is now an interactive — and effective — way to select students who will most benefit from its programs. The local schools’ curricula, Hassan says, might not provide enough of an academic challenge. “We can’t teach everyone and accommodate everyone because there are a lot of schools,” Hassan says, “but we offer another place for knowledge.” 

      The hackathon helps students develop data science and robotics skills. Before they start coding, students have to convince ADAI Circle teachers that their designs are viable, answering questions like, “What problem are you solving?” and “How will this help the community?” A community-oriented mindset is just as important to the curriculum.

      In addition to the practical skills Hassan gained from Emerging Talent, he leveraged the program’s network to help his community. Thanks to a social media connection Hassan made with the nongovernmental organization Give Internet after one of Emerging Talent’s virtual events, Give Internet brought internet access to ADAI Circle.

      Bridging the AI gap to unmet communities

      In 2023, ADAI Circle connected with another MIT Open Learning program, Responsible AI for Social Empowerment and Education (RAISE), which led to a pilot test of a project-based AI curriculum for middle school students. The Responsible AI for Computational Action (RAICA) curriculum equipped ADAI Circle students with AI skills for chatbots and natural language processing. 

      “I liked that program because it was based on what we’re teaching at the center,” Hassan says, speaking of his organization’s mission of bridging the AI gap to reach unmet communities.

      The RAICA curriculum was designed by education experts at MIT Scheller Teacher Education Program (STEP Lab) and AI experts from MIT Personal Robots group and MIT App Inventor. ADAI Circle teachers gave detailed feedback about the pilot to the RAICA team. During weekly meetings with Glenda Stump, education research scientist for RAICA and J-WEL, and Angela Daniel, teacher development specialist for RAICA, the teachers discussed their experiences, prepared for upcoming lessons, and translated the learning materials in real time. 

      “We are trying to create a curriculum that’s accessible worldwide and to students who typically have little or no access to technology,” says Mary Cate Gustafson-Quiett, curriculum design manager at STEP Lab and project manager for RAICA. “Working with ADAI and students in a refugee camp challenged us to design in more culturally and technologically inclusive ways.”

      Gustafson-Quiett says the curriculum feedback from ADAI Circle helped inform how RAICA delivers teacher development resources to accommodate learning environments with limited internet access. “They also exposed places where our team’s western ideals, specifically around individualism, crept into activities in the lesson and contrasted with their more communal cultural beliefs,” she says.

      Eager to introduce more MIT-developed AI resources, Hassan also shared MIT RAISE’s Day of AI curricula with ADAI Circle teachers. The new ChatGPT module gave students the chance to level up their chatbot programming skills that they gained from the RAICA module. Some of the advanced students are taking initiative to use ChatGPT API to create their own projects in education.

      “We don’t want to tell them what to do, we want them to come up with their own ideas,” Hassan says.

      Although ADAI Circle faces many challenges, Hassan says his team is addressing them one by one. Last year, they didn’t have electricity in their Innovation Hub, but they solved that. This year, they achieved a stable internet connection that’s one of the fastest in Malawi. Next up, they are hoping to secure more devices for their students, create more jobs, and add additional hubs throughout the community. The work is never done, but Hassan is starting to see the impact that ADAI Circle is making. 

      “For those who want to learn data science, let’s let them learn,” Hassan says. More

    • 63 Shares159 Views

      in Data Management & Statistics

      Leveraging language to understand machines

      by

      Natural language conveys ideas, actions, information, and intent through context and syntax; further, there are volumes of it contained in databases. This makes it an excellent source of data to train machine-learning systems on. Two master’s of engineering students in the 6A MEng Thesis Program at MIT, Irene Terpstra ’23 and Rujul Gandhi ’22, are working with mentors in the MIT-IBM Watson AI Lab to use this power of natural language to build AI systems.

      As computing is becoming more advanced, researchers are looking to improve the hardware that they run on; this means innovating to create new computer chips. And, since there is literature already available on modifications that can be made to achieve certain parameters and performance, Terpstra and her mentors and advisors Anantha Chandrakasan, MIT School of Engineering dean and the Vannevar Bush Professor of Electrical Engineering and Computer Science, and IBM’s researcher Xin Zhang, are developing an AI algorithm that assists in chip design.

      “I’m creating a workflow to systematically analyze how these language models can help the circuit design process. What reasoning powers do they have, and how can it be integrated into the chip design process?” says Terpstra. “And then on the other side, if that proves to be useful enough, [we’ll] see if they can automatically design the chips themselves, attaching it to a reinforcement learning algorithm.”

      To do this, Terpstra’s team is creating an AI system that can iterate on different designs. It means experimenting with various pre-trained large language models (like ChatGPT, Llama 2, and Bard), using an open-source circuit simulator language called NGspice, which has the parameters of the chip in code form, and a reinforcement learning algorithm. With text prompts, researchers will be able to query how the physical chip should be modified to achieve a certain goal in the language model and produced guidance for adjustments. This is then transferred into a reinforcement learning algorithm that updates the circuit design and outputs new physical parameters of the chip.

      “The final goal would be to combine the reasoning powers and the knowledge base that is baked into these large language models and combine that with the optimization power of the reinforcement learning algorithms and have that design the chip itself,” says Terpstra.

      Rujul Gandhi works with the raw language itself. As an undergraduate at MIT, Gandhi explored linguistics and computer sciences, putting them together in her MEng work. “I’ve been interested in communication, both between just humans and between humans and computers,” Gandhi says.

      Robots or other interactive AI systems are one area where communication needs to be understood by both humans and machines. Researchers often write instructions for robots using formal logic. This helps ensure that commands are being followed safely and as intended, but formal logic can be difficult for users to understand, while natural language comes easily. To ensure this smooth communication, Gandhi and her advisors Yang Zhang of IBM and MIT assistant professor Chuchu Fan are building a parser that converts natural language instructions into a machine-friendly form. Leveraging the linguistic structure encoded by the pre-trained encoder-decoder model T5, and a dataset of annotated, basic English commands for performing certain tasks, Gandhi’s system identifies the smallest logical units, or atomic propositions, which are present in a given instruction.

      “Once you’ve given your instruction, the model identifies all the smaller sub-tasks you want it to carry out,” Gandhi says. “Then, using a large language model, each sub-task can be compared against the available actions and objects in the robot’s world, and if any sub-task can’t be carried out because a certain object is not recognized, or an action is not possible, the system can stop right there to ask the user for help.”

      This approach of breaking instructions into sub-tasks also allows her system to understand logical dependencies expressed in English, like, “do task X until event Y happens.” Gandhi uses a dataset of step-by-step instructions across robot task domains like navigation and manipulation, with a focus on household tasks. Using data that are written just the way humans would talk to each other has many advantages, she says, because it means a user can be more flexible about how they phrase their instructions.

      Another of Gandhi’s projects involves developing speech models. In the context of speech recognition, some languages are considered “low resource” since they might not have a lot of transcribed speech available, or might not have a written form at all. “One of the reasons I applied to this internship at the MIT-IBM Watson AI Lab was an interest in language processing for low-resource languages,” she says. “A lot of language models today are very data-driven, and when it’s not that easy to acquire all of that data, that’s when you need to use the limited data efficiently.” 

      Speech is just a stream of sound waves, but humans having a conversation can easily figure out where words and thoughts start and end. In speech processing, both humans and language models use their existing vocabulary to recognize word boundaries and understand the meaning. In low- or no-resource languages, a written vocabulary might not exist at all, so researchers can’t provide one to the model. Instead, the model can make note of what sound sequences occur together more frequently than others, and infer that those might be individual words or concepts. In Gandhi’s research group, these inferred words are then collected into a pseudo-vocabulary that serves as a labeling method for the low-resource language, creating labeled data for further applications.

      The applications for language technology are “pretty much everywhere,” Gandhi says. “You could imagine people being able to interact with software and devices in their native language, their native dialect. You could imagine improving all the voice assistants that we use. You could imagine it being used for translation or interpretation.” More

    • 138 Shares169 Views

      in Data Management & Statistics

      Automated system teaches users when to collaborate with an AI assistant

      by

      Artificial intelligence models that pick out patterns in images can often do so better than human eyes — but not always. If a radiologist is using an AI model to help her determine whether a patient’s X-rays show signs of pneumonia, when should she trust the model’s advice and when should she ignore it?

      A customized onboarding process could help this radiologist answer that question, according to researchers at MIT and the MIT-IBM Watson AI Lab. They designed a system that teaches a user when to collaborate with an AI assistant.

      In this case, the training method might find situations where the radiologist trusts the model’s advice — except she shouldn’t because the model is wrong. The system automatically learns rules for how she should collaborate with the AI, and describes them with natural language.

      During onboarding, the radiologist practices collaborating with the AI using training exercises based on these rules, receiving feedback about her performance and the AI’s performance.

      The researchers found that this onboarding procedure led to about a 5 percent improvement in accuracy when humans and AI collaborated on an image prediction task. Their results also show that just telling the user when to trust the AI, without training, led to worse performance.

      Importantly, the researchers’ system is fully automated, so it learns to create the onboarding process based on data from the human and AI performing a specific task. It can also adapt to different tasks, so it can be scaled up and used in many situations where humans and AI models work together, such as in social media content moderation, writing, and programming.

      “So often, people are given these AI tools to use without any training to help them figure out when it is going to be helpful. That’s not what we do with nearly every other tool that people use — there is almost always some kind of tutorial that comes with it. But for AI, this seems to be missing. We are trying to tackle this problem from a methodological and behavioral perspective,” says Hussein Mozannar, a graduate student in the Social and Engineering Systems doctoral program within the Institute for Data, Systems, and Society (IDSS) and lead author of a paper about this training process.

      The researchers envision that such onboarding will be a crucial part of training for medical professionals.

      “One could imagine, for example, that doctors making treatment decisions with the help of AI will first have to do training similar to what we propose. We may need to rethink everything from continuing medical education to the way clinical trials are designed,” says senior author David Sontag, a professor of EECS, a member of the MIT-IBM Watson AI Lab and the MIT Jameel Clinic, and the leader of the Clinical Machine Learning Group of the Computer Science and Artificial Intelligence Laboratory (CSAIL).

      Mozannar, who is also a researcher with the Clinical Machine Learning Group, is joined on the paper by Jimin J. Lee, an undergraduate in electrical engineering and computer science; Dennis Wei, a senior research scientist at IBM Research; and Prasanna Sattigeri and Subhro Das, research staff members at the MIT-IBM Watson AI Lab. The paper will be presented at the Conference on Neural Information Processing Systems.

      Training that evolves

      Existing onboarding methods for human-AI collaboration are often composed of training materials produced by human experts for specific use cases, making them difficult to scale up. Some related techniques rely on explanations, where the AI tells the user its confidence in each decision, but research has shown that explanations are rarely helpful, Mozannar says.

      “The AI model’s capabilities are constantly evolving, so the use cases where the human could potentially benefit from it are growing over time. At the same time, the user’s perception of the model continues changing. So, we need a training procedure that also evolves over time,” he adds.

      To accomplish this, their onboarding method is automatically learned from data. It is built from a dataset that contains many instances of a task, such as detecting the presence of a traffic light from a blurry image.

      The system’s first step is to collect data on the human and AI performing this task. In this case, the human would try to predict, with the help of AI, whether blurry images contain traffic lights.

      The system embeds these data points onto a latent space, which is a representation of data in which similar data points are closer together. It uses an algorithm to discover regions of this space where the human collaborates incorrectly with the AI. These regions capture instances where the human trusted the AI’s prediction but the prediction was wrong, and vice versa.

      Perhaps the human mistakenly trusts the AI when images show a highway at night.

      After discovering the regions, a second algorithm utilizes a large language model to describe each region as a rule, using natural language. The algorithm iteratively fine-tunes that rule by finding contrasting examples. It might describe this region as “ignore AI when it is a highway during the night.”

      These rules are used to build training exercises. The onboarding system shows an example to the human, in this case a blurry highway scene at night, as well as the AI’s prediction, and asks the user if the image shows traffic lights. The user can answer yes, no, or use the AI’s prediction.

      If the human is wrong, they are shown the correct answer and performance statistics for the human and AI on these instances of the task. The system does this for each region, and at the end of the training process, repeats the exercises the human got wrong.

      “After that, the human has learned something about these regions that we hope they will take away in the future to make more accurate predictions,” Mozannar says.

      Onboarding boosts accuracy

      The researchers tested this system with users on two tasks — detecting traffic lights in blurry images and answering multiple choice questions from many domains (such as biology, philosophy, computer science, etc.).

      They first showed users a card with information about the AI model, how it was trained, and a breakdown of its performance on broad categories. Users were split into five groups: Some were only shown the card, some went through the researchers’ onboarding procedure, some went through a baseline onboarding procedure, some went through the researchers’ onboarding procedure and were given recommendations of when they should or should not trust the AI, and others were only given the recommendations.

      Only the researchers’ onboarding procedure without recommendations improved users’ accuracy significantly, boosting their performance on the traffic light prediction task by about 5 percent without slowing them down. However, onboarding was not as effective for the question-answering task. The researchers believe this is because the AI model, ChatGPT, provided explanations with each answer that convey whether it should be trusted.

      But providing recommendations without onboarding had the opposite effect — users not only performed worse, they took more time to make predictions.

      “When you only give someone recommendations, it seems like they get confused and don’t know what to do. It derails their process. People also don’t like being told what to do, so that is a factor as well,” Mozannar says.

      Providing recommendations alone could harm the user if those recommendations are wrong, he adds. With onboarding, on the other hand, the biggest limitation is the amount of available data. If there aren’t enough data, the onboarding stage won’t be as effective, he says.

      In the future, he and his collaborators want to conduct larger studies to evaluate the short- and long-term effects of onboarding. They also want to leverage unlabeled data for the onboarding process, and find methods to effectively reduce the number of regions without omitting important examples.

      “People are adopting AI systems willy-nilly, and indeed AI offers great potential, but these AI agents still sometimes makes mistakes. Thus, it’s crucial for AI developers to devise methods that help humans know when it’s safe to rely on the AI’s suggestions,” says Dan Weld, professor emeritus at the Paul G. Allen School of Computer Science and Engineering at the University of Washington, who was not involved with this research. “Mozannar et al. have created an innovative method for identifying situations where the AI is trustworthy, and (importantly) to describe them to people in a way that leads to better human-AI team interactions.”

      This work is funded, in part, by the MIT-IBM Watson AI Lab. More