arXiv subjects

Jessie Liu

Publications and source records attributed to Jessie Liu.

9 recordsLinked to original sources

arXiv2025

Spiral of Silence: How Neutral Moderation Polarizes Content Creation

This paper investigates how content moderation affects content creation in an ideologically diverse online environments. We develop a model in which users act as both creators and consumers, differing in their ideological affiliation and propensity to produce toxic content. Affective polarization, i.e., users' aversion to ideologically opposed content, interacts with moderation in unintended ways. We show that even ideologically neutral moderation that targets only toxicity can suppress non-toxic content creation, particularly from ideological minorities. Our analysis reveals a content-level externality: when toxic content is removed, non-toxic posts gain exposure. While creators from the ideological majority group sometimes benefit from this exposure, they do not internalize the negative spillovers, i.e., increased out-group animosity toward minority creators. This can discourage minority creation and polarize the content supply, ultimately leaving minority users in a more ideologically imbalanced environment: a mechanism reminiscent of the "spiral of silence." Thus, our model offers an alternative perspective to a common debate: what appears as bias in moderation needs not reflect bias in rules, but can instead emerge endogenously as self-censorship in equilibrium. We also extend the model to explore how content personalization interacts with moderation policies.

econ.GN↗

arXiv2025

Reactive Marketing and the Co-Production of (In)Authenticity

Businesses often react to external events by sending pro-social messages on social media that show the sender's alignment with the underlying prosocial cause and enhance their brand image. Consumers are uncertain about the authenticity of such messages because a company can choose to send prosocial messages even when their alignment with the social cause is not genuine. We study the sender's incentives to send (in)authentic messages and the consumer's reactions when an external investigator can verify the sender's message. We find that the sender's equilibrium strategy depends on the receiver's emphasis on external investigation versus their self-signaling incentives. When the receiver is more internally focused, the sender chooses self-sufficiency strategy, building credibility independent of external investigation. When the receiver is more externally focused, the sender can either use self-sufficiency or complementarity, the latter relying on validation by the external investigator. Thus, authenticity is coproduced by the sender, the receiver, and the investigator. Importantly, self-sufficiency results in more authenticity in reactive marketing messages than complementarity. We extend the model to incorporate confirmation bias of the receiver. We show that confirmation bias can act a doubled-edged sword, sometimes making the sender's persuasion task easier and other times making it more difficult.

econ.TH↗

arXiv2025

The 2025 OpenAI Preparedness Framework does not guarantee any AI risk mitigation practices: a proof-of-concept for affordance analyses of AI safety policies

Prominent AI companies are producing 'safety frameworks' as a type of voluntary self-governance. These statements purport to establish risk thresholds and safety procedures for the development and deployment of highly capable AI. Understanding which AI risks are covered and what actions are allowed, refused, demanded, encouraged, or discouraged by these statements is vital for assessing how these frameworks actually govern AI development and deployment. We draw on affordance theory to analyse the OpenAI 'Preparedness Framework Version 2' (April 2025) using the Mechanisms & Conditions model of affordances and the MIT AI Risk Repository. We find that this safety policy requests evaluation of a small minority of AI risks, encourages deployment of systems with 'Medium' capabilities for unintentionally enabling 'severe harm' (which OpenAI defines as >1000 deaths or >$100B in damages), and allows OpenAI's CEO to deploy even more dangerous capabilities. These findings suggest that effective mitigation of AI risks requires more robust governance interventions beyond current industry self-regulation. Our affordance analysis provides a replicable method for evaluating what safety frameworks actually permit versus what they claim.

cs.CY↗

arXiv2025

Techniques for improved statistical convergence in quantification of eddy diffusivity moments

While recent approaches, such as the macroscopic forcing method (MFM) or Green's function-based approaches, can be used to compute Reynolds-averaged Navier--Stokes closure operators using forced direct numerical simulations, MFM can also be used to directly compute moments of the effective nonlocal and anisotropic eddy diffusivities. The low-order spatial and temporal moments contain limited information about the eddy diffusivity but are often sufficient for quantification and modeling of nonlocal and anisotropic effects. However, when using MFM to compute eddy diffusivity moments, the statistical convergence can be slow for higher-order moments. In this work, we demonstrate that using the same direct numerical simulation (DNS) for all forced MFM simulations improves statistical convergence of the eddy diffusivity moments. We present its implementation in conjunction with a decomposition method that handles the MFM forcing semi-analytically and allows for consistent boundary condition treatment, which we develop for both scalar and momentum transport. We demonstrate that for a two-dimensional Rayleigh--Taylor instability case study, using the same DNS for all forced MFM simulations results in convergence with O(100) simulations rather than O(1000) simulations. We then demonstrate the impacts of improved convergence on the quantification of the eddy diffusivity.

physics.flu-dyn↗

arXiv2023

Targeted computation of nonlocal closure operators via an adjoint-based macroscopic forcing method

Reynolds-averaged Navier--Stokes (RANS) closure must be sensitive to the flow physics, including nonlocality and anisotropy of the effective eddy viscosity. Recent approaches used forced direct numerical simulations to probe these effects, including the macroscopic forcing method (MFM) of Mani and Park ($\textit{Phys. Rev. Fluids}$ $\textbf{6}$, 054607 (2021)) and the Green's function approach of Hamba ($\textit{Phys. Fluids}$ $\textbf{17}$, 115102 (2005)). The resulting nonlocal and anisotropic eddy viscosities are exact and relate Reynolds stresses to mean velocity gradients at all locations. They can be used to inform RANS models of the sensitivity to the mean velocity gradient and the suitability of local and isotropic approximations. However, these brute-force approaches are expensive. They force the mean velocity gradient at each point in the averaged space and measure the Reynolds stress response, requiring a separate simulation for each mean velocity gradient location. Thus, computing the eddy viscosity requires as many simulations as degrees of freedom in the averaged space, which can be cost-prohibitive for problems with many degrees of freedom. In this work, we develop an adjoint-based MFM to obtain the eddy viscosity at a given Reynolds stress location using a single simulation. This approach recovers the Reynolds stress dependence at a location of interest, such as a separation point or near a wall, on the mean velocity gradient at all locations. We demonstrate using adjoint MFM to compute the eddy viscosity for a specified wall-normal location in an incompressible turbulent channel flow using one simulation. In contrast, a brute-force approach for the same problem requires $N=144$ simulations (the number of grid points in the non-averaged coordinate direction). We show that a local approximation for the eddy viscosity would have been inappropriate.

physics.flu-dyn↗

arXiv2023

Nonlocality of Mean Scalar Transport in Two-Dimensional Rayleigh-Taylor Instability Using the Macroscopic Forcing Method

The importance of nonlocality of mean scalar transport in 2D Rayleigh-Taylor Instability (RTI) is investigated. The Macroscopic Forcing Method (MFM) is utilized to measure spatio-temporal moments of the eddy diffusivity kernel representing passive scalar transport in the ensemble averaged fields. Presented in this work are several studies assessing the importance of the higher-order moments of the eddy diffusivity, which contain information about nonlocality, in models for RTI. First, it is demonstrated through a comparison of leading-order models that a purely local eddy diffusivity is insufficient in capturing the mean field evolution of the mass fraction in RTI. Therefore, higher-order moments of the eddy diffusivity operator are not negligible. Models are then constructed by utilizing the measured higher-order moments. It is demonstrated that an explicit operator based on the Kramers-Moyal expansion of the eddy diffusivity kernel is insufficient. An implicit operator construction that matches the measured moments is shown to offer improvements relative to the local model in a converging fashion.

physics.flu-dyn↗

arXiv2023

A systematic approach for modeling a nonlocal eddy diffusivity

This study considers advective and diffusive transport of passive scalar fields by spatially-varying incompressible flows. Prior studies have shown that the eddy diffusivities governing the mean field transport in such systems can generally be nonlocal in space and time. While for many flows nonlocal eddy diffusivities are more accurate than commonly-used Boussinesq eddy diffusivities, nonlocal eddy diffusivities are often computationally cost-prohibitive to obtain and difficult to implement in practice. We develop a systematic and more cost-effective approach for modeling nonlocal eddy diffusivities using matched moment inverse (MMI) operators. These operators are constructed using only a few leading-order moments of the exact nonlocal eddy diffusivity kernel, which can be easily computed using the inverse macroscopic forcing method (IMFM) (Mani and Park (2021)). The resulting reduced-order models for the mean fields that incorporate the modeled eddy diffusivities often improve Boussinesq-limit models since they capture leading-order nonlocal effects. But more importantly, these models can be expressed as partial differential equations that are readily solvable using existing computational fluid dynamics capabilities rather than as integro-partial differential equations.

physics.flu-dyn↗

arXiv2023

Fast Macroscopic Forcing Method

The macroscopic forcing method (MFM) of Mani and Park and similar methods for obtaining turbulence closure operators, such as the Green's function-based approach of Hamba, recover reduced solution operators from repeated direct numerical simulations (DNS). MFM has been used to quantify RANS-like operators for homogeneous isotropic turbulence and turbulent channel flows. Standard algorithms for MFM force each coarse-scale degree of freedom (i.e., degree of freedom in the RANS space) and conduct a corresponding fine-scale simulation (i.e., DNS), which is expensive. We combine this method with an approach recently proposed by Schäfer and Owhadi (2023) to recover elliptic integral operators from a polylogarithmic number of matrix-vector products. The resulting Fast MFM introduced in this work applies sparse reconstruction to expose local features in the closure operator and reconstructs this coarse-grained differential operator in only a few matrix-vector products and correspondingly, a few MFM simulations. For flows with significant nonlocality, the algorithm first "peels" long-range effects with dense matrix-vector products to expose a local operator. We demonstrate the algorithm's performance for scalar transport in a laminar channel flow and momentum transport in a turbulent one. For these, we recover eddy diffusivity operators at 1% of the cost of computing the exact operator via a brute-force approach for the laminar channel flow problem and 13% for the turbulent one. We observe that we can reconstruct these operators with an increase in accuracy by about a factor of 100 over randomized low-rank methods. We glean that for problems in which the RANS space is reducible to one dimension, eddy diffusivity and eddy viscosity operators can be reconstructed with reasonable accuracy using only a few simulations, regardless of simulation resolution or degrees of freedom.

physics.comp-ph↗

arXiv2021

Incorporating Uncertainty in Learning to Defer Algorithms for Safe Computer-Aided Diagnosis

Deep neural networks are increasingly being used for computer-aided diagnosis, but erroneous diagnoses can be extremely costly for patients. We propose a learning to defer with uncertainty (LDU) algorithm which identifies patients for whom diagnostic uncertainty is high and defers them for evaluation by human experts. LDU was evaluated on the diagnosis of myocardial infarction (using discharge summaries), the diagnosis of any comorbidities (using structured data), and the diagnosis of pleural effusion and pneumothorax (using chest x-rays), and compared with 'learning to defer without uncertainty information' (LD) and 'direct triage by uncertainty' (DT) methods. LDU achieved the same F1 score as LD but deferred considerably fewer patients (e.g. 36% vs. 69% deferral rate for diagnosing pleural effusion with an F1 score of 0.96). Furthermore, even when many patients were assigned the wrong diagnosis with high confidence (e.g. for the diagnosis of any comorbidities) LDU achieved a 17% increase in F1 score, whereas DT was not applicable. Importantly, the weight of the defer loss in LDU can be easily adjusted to obtain the desired trade-off between diagnostic accuracy and deferral rate. In conclusion, LDU can readily augment any existing diagnostic network to reduce the risk of erroneous diagnoses in clinical practice.

cs.LG↗