Researcher profile

Douglas Zytko

Douglas Zytko contributes to research discovery and scholarly infrastructure.

ResearcherAffiliation not importedOpen to collaborate

Trust snapshot

Quick read

Trust 21 - EmergingVerification L1Unclaimed author
6works
0followers
3topics
4close collaborators

Actions

Decide how to stay connected

Follow researcher0

Identity and collaboration

How to connect with this researcher

Claiming links this public author record to a researcher profile and unlocks direct collaboration workflows.

Log in to claim

Direct collaboration

Open a focused conversation when the fit is right

Claim this author entity first to unlock direct invitations.

Research graph

See the researcher in context

Open full explorer

Inspect adjacent work, topics, institutions and collaborators without jumping out to a separate graph page.

Building this graph slice

BZPEER is loading the nearby papers, people, topics and institutions for this page.

Published work

6 published item(s)

preprint2025arXiv

Not All Tokens Are Meant to Be Forgotten

Large Language Models (LLMs), pre-trained on massive text corpora, exhibit remarkable human-level language understanding, reasoning, and decision-making abilities. However, they tend to memorize unwanted information, such as private or copyrighted content, raising significant privacy and legal concerns. Unlearning has emerged as a promising solution, but existing methods face a significant challenge of over-forgetting. This issue arises because they indiscriminately suppress the generation of all the tokens in forget samples, leading to a substantial loss of model utility. To overcome this challenge, we introduce the Targeted Information Forgetting (TIF) framework, which consists of (1) a flexible targeted information identifier designed to differentiate between unwanted words (UW) and general words (GW) in the forget samples, and (2) a novel Targeted Preference Optimization approach that leverages Logit Preference Loss to unlearn unwanted information associated with UW and Preservation Loss to retain general information in GW, effectively improving the unlearning process while mitigating utility degradation. Extensive experiments on the TOFU and MUSE benchmarks demonstrate that the proposed TIF framework enhances unlearning effectiveness while preserving model utility and achieving state-of-the-art results.

preprint2022arXiv

Designing AI for Online-to-Offline Safety Risks with Young Women: The Context of Social Matching

In this position paper we draw attention to safety risks against youth and young adults that originate through the combination of online and in-person interaction, and opportunities for AI to address these risks. Our context of study is social matching systems (e.g., Tinder, Bumble), which are used by young adults for online-to-offline interaction with strangers, and which are correlated with sexual violence both online and in-person. The paper presents early insights from an ongoing participatory AI design study in which young women build directly explainable models for detecting risk associated with discovered social opportunities, and articulate what AI should do once risk has been detected. We seek to advocate for participatory AI design as a way to directly incorporate youth and young adults into the design of a safer Internet. We also draw attention to challenges with the method.

preprint2022arXiv

Designing Social VR: A Collection of Design Choices Across Commercial and Research Applications

Social VR has experienced tremendous growth in the commercial space recently as an emerging technology for rich interactions themed around leisure, work, and relationship building. As a result, the state of social VR application design has become rapidly obfuscated, which complicates identification of design trends and uncommon features that could inform future design, and hinders inclusion of new voices in this design space. To help address this problem, we present a taxonomy of social VR application design choices as informed by 44 commercial and prototypical applications. Our taxonomy was informed by multiple discovery strategies including literature review, search of VR-themed subreddits, and autobiographical landscape research. The taxonomy elucidates various features across three design areas: the self, interaction, and the environment.

preprint2022arXiv

Ethical Considerations When Constructing Participatory Design Protocols for Social Robots

Participatory design has emerged as a popular approach to foreground ethical considerations in social robots by incorporating anticipated users and stakeholders as designers. Here we draw attention to the ethics of participatory design as a method, distinct from the ethical considerations of the social robot being co-designed. More specifically, we consider the ethical concerns posed by the act of stakeholder participation - the morals and values that should be explicitly considered when we, as researchers or practitioners, devise protocols for participatory design of social robots ("how" stakeholders participate). We use the case of robot-assisted sexual violence mitigation to exemplify ethical considerations of participatory design protocols such as risk of harm, exploitation, and reduction of stakeholder agency. To incorporate these and other ethical considerations in the creation of social robot participatory design protocols, we advocate letting stakeholders design their own form of participation by including them in the creation of participatory design sessions, structures, and processes.

preprint2022arXiv

Human-AI Interaction for User Safety in Social Matching Apps: Involving Marginalized Users in Design

In this position paper we intend to advocate for participatory design methods and mobile social matching apps as ripe contexts for exploring novel human-AI interactions that benefit marginalized groups. Mobile social matching apps like Tinder and Bumble use AI to introduce users to each other for rapid face-to-face meetings. These user discoveries and subsequent interactions pose disproportionate risk of sexual violence and other harms to marginalized user demographics, specifically women and the LGBTQIA+ community. We want to extend the role of AI in these apps to keep users safe while they interact with strangers across online and offline modalities. To do this, we are using participatory design methods to empower women and LGBTQIA+ individuals to envision future human-AI interactions that prioritize their safety during social matching app-use. In one study, stakeholders identifying as LGBTQIA+ or women are redesigning dating apps to mediate exchange of sexual consent and therefore mitigate sexual violence. In the other study, women are designing multi-purpose, opportunistic social matching apps that foreground women's safety.

preprint2022arXiv

The...Tinderverse?: Opportunities and Challenges for User Safety in Extended Reality (XR) Dating Apps

Dating apps such as Tinder have announced plans for a dating metaverse: the incorporation of XR technologies into the online dating process to augment interactions between potential sexual partners across virtual and physical worlds. While the dating metaverse is still in conceptual stages we can forecast significant harms that it may expose daters to given prior research into the frequency and severity of sexual harms facilitated by dating apps as well as harms within social VR environments. In this workshop paper we envision how XR could enrich virtual-to-physical interaction between potential sexual partners and outline harms that it will likely perpetuate as well. We then introduce our ongoing research to preempt such harms: a participatory design study with sexual violence experts and demographics at disproportionate risk of sexual violence to produce mitigative solutions to sexual violence perpetuated by XR-enabled dating apps.