arXiv subjects

Michelle Feng

Publications and source records attributed to Michelle Feng.

8 recordsLinked to original sources

arXiv2023

The Impact of Imperfect XAI on Human-AI Decision-Making

Explainability techniques are rapidly being developed to improve human-AI decision-making across various cooperative work settings. Consequently, previous research has evaluated how decision-makers collaborate with imperfect AI by investigating appropriate reliance and task performance with the aim of designing more human-centered computer-supported collaborative tools. Several human-centered explainable AI (XAI) techniques have been proposed in hopes of improving decision-makers' collaboration with AI; however, these techniques are grounded in findings from previous studies that primarily focus on the impact of incorrect AI advice. Few studies acknowledge the possibility of the explanations being incorrect even if the AI advice is correct. Thus, it is crucial to understand how imperfect XAI affects human-AI decision-making. In this work, we contribute a robust, mixed-methods user study with 136 participants to evaluate how incorrect explanations influence humans' decision-making behavior in a bird species identification task, taking into account their level of expertise and an explanation's level of assertiveness. Our findings reveal the influence of imperfect XAI and humans' level of expertise on their reliance on AI and human-AI team performance. We also discuss how explanations can deceive decision-makers during human-AI collaboration. Hence, we shed light on the impacts of imperfect XAI in the field of computer-supported cooperative work and provide guidelines for designers of human-AI collaboration systems.

cs.HC↗

arXiv2023

Repulsive Bounded-Confidence Model of Opinion Dynamics in Polarized Communities

Collective opinions affect civic participation, governance, and societal norms. Due to the influence of opinion dynamics, many models of their formation and evolution have been developed. A commonly used approach for the study of opinion dynamics is bounded-confidence models. In these models, individuals are influenced by the opinions of others in their network. They generally assume that individuals will formulate their opinions to resemble those of their peers. In this paper, inspired by the dynamics of partisan politics, we introduce a bounded-confidence model in which individuals may be repelled by the opinions of their peers rather than only attracted to them. We prove convergence properties of our model and perform simulations to study the behavior of our model on various types of random networks. In particular, we observe that including opinion repulsion leads to a higher degree of opinion fragmentation than in standard bounded-confidence models.

math.DS↗

arXiv2021

An Adaptive Bounded-Confidence Model of Opinion Dynamics on Networks

Individuals who interact with each other in social networks often exchange ideas and influence each other's opinions. A popular approach to study the spread of opinions on networks is by examining bounded-confidence models (BCMs), in which the nodes of a network have continuous-valued states that encode their opinions and are receptive to other nodes' opinions when they lie within some confidence bound of their own opinion. In this paper, we extend the Deffuant--Weisbuch (DW) model, which is a well-known BCM, by examining the spread of opinions that coevolve with network structure. We propose an adaptive variant of the DW model in which the nodes of a network can (1) alter their opinions when they interact with neighboring nodes and (2) break connections with neighbors based on an opinion tolerance threshold and then form new connections following the principle of homophily. This opinion tolerance threshold determines whether or not the opinions of adjacent nodes are sufficiently different to be viewed as `discordant'. Using numerical simulations, we find that our adaptive DW model requires a larger confidence bound than a baseline DW model for the nodes of a network to achieve a consensus opinion. In one region of parameter space, we observe `pseudo-consensus' steady states, in which there exist multiple subclusters of an opinion cluster with opinions that differ from each other by a small amount. In our simulations, we also examine the importance of early-time dynamics and nodes with initially moderate opinions for achieving consensus. Additionally, we explore the effects of coevolution on the convergence time of our BCM.

physics.soc-ph↗

arXiv2021

Topological Data Analysis of Spatial Systems

In this chapter, we discuss applications of topological data analysis (TDA) to spatial systems. We briefly review the recently proposed level-set construction of filtered simplicial complexes, and we then examine persistent homology in two cases studies: street networks in Shanghai and hotspots of COVID-19 infections. We then summarize our results and provide an outlook on TDA in spatial systems.

cs.SI↗

arXiv2021

A Bounded-Confidence Model of Opinion Dynamics on Hypergraphs

People's opinions evolve over time as they interact with their friends, family, colleagues, and others. In the study of opinion dynamics on networks, one often encodes interactions between people in the form of dyadic relationships, but many social interactions in real life are polyadic (i.e., they involve three or more people). In this paper, we extend an asynchronous bounded-confidence model (BCM) on graphs, in which nodes are connected pairwise by edges, to an asynchronous BCM on hypergraphs, in which arbitrarily many nodes can be connected by a single hyperedge. We show that our hypergraph BCM converges to consensus under a wide range of initial conditions for the opinions of the nodes, including for non-uniform and asymmetric initial opinion distributions. We also show that, under suitable conditions, echo chambers can form on hypergraphs with community structure. We demonstrate that the opinions of individuals can sometimes jump from one opinion cluster to another in a single time step, a phenomenon (which we call ``opinion jumping'') that is not possible in standard dyadic BCMs. Additionally, we observe that there is a phase transition in the convergence time on {a complete hypergraph} when the variance $\sigma^2$ of the initial opinion distribution equals the confidence bound $c$. We prove that the convergence time grows at least exponentially fast with the number of nodes when $\sigma^2 > c$ and the initial opinions are normally distributed. Therefore, to determine the convergence properties of our hypergraph BCM when the variance and the number of hyperedges are both large, it is necessary to use analytical methods instead of relying only on Monte Carlo simulations.

cs.SI↗

arXiv2020

Connecting the Dots: Discovering the "Shape" of Data

Scientists use a mathematical subject called 'topology' to study the shapes of objects. An important part of topology is counting the numbers of pieces and holes in objects, and people use this information to group objects into different types. For example, a doughnut has the same number of holes and the same number of pieces as a teacup with one handle, but it is different from a ball. In studies that resemble activities like "connect the dots", scientists use ideas from topology to study the shape of data. Data can take many possible forms: a picture made of dots, a large collection of numbers from a scientific experiment, or something else. The approach in these studies is called 'topological data analysis', and it has been used to study the branching structures of veins in leaves, how people vote in elections, flight patterns in models of bird flocking, and more. Scientists can take data on the way veins branch on leaves and use topological data analysis to divide the leaves into different groups and discover patterns that may otherwise be hard to find.

math.HO↗

arXiv2020

Spatial Applications of Topological Data Analysis: Cities, Snowflakes, Random Structures, and Spiders Spinning Under the Influence

Spatial networks are ubiquitous in social, geographical, physical, and biological applications. To understand the large-scale structure of networks, it is important to develop methods that allow one to directly probe the effects of space on structure and dynamics. Historically, algebraic topology has provided one framework for rigorously and quantitatively describing the global structure of a space, and recent advances in topological data analysis (TDA) have given scholars a new lens for analyzing network data. In this paper, we study a variety of spatial networks -- including both synthetic and natural ones -- using novel topological methods that we recently developed for analyzing spatial networks. We demonstrate that our methods are able to capture meaningful quantities, with specifics that depend on context, in spatial networks and thereby provide useful insights into the structure of those networks, including a novel approach for characterizing them based on their topological structures. We illustrate these ideas with examples of synthetic networks and dynamics on them, street networks in cities, snowflakes, and webs spun by spiders under the influence of various psychotropic substances.

cs.SI↗

arXiv2019

Persistent Homology of Geospatial Data: A Case Study with Voting

A crucial step in the analysis of persistent homology is the transformation of data into an appropriate topological object (in our case, a simplicial complex). Modern packages for persistent homology often construct Vietoris--Rips or other distance-based simplicial complexes on point clouds because they are relatively easy to compute. We investigate alternative methods of constructing these complexes and the effects of making associated choices during simplicial-complex construction on the output of persistent-homology algorithms. We present two new methods for constructing simplicial complexes from two-dimensional geospatial data (such as maps). We apply these methods to a California precinct-level voting data set, demonstrating that our new constructions can capture geometric characteristics that are missed by distance-based constructions. Our new constructions can thus yield more interpretable persistence modules and barcodes for geospatial data. In particular, they are able to distinguish short-persistence features that occur only for a narrow range of distance scales (e.g., voting behaviors in densely populated cities) from short-persistence noise by incorporating information about other spatial relationships between precincts.

cs.CG↗