Anonymous Interactions Research Articles

SecureVFL: privacy-preserving multi-party vertical federated learning based on blockchain and RSS

Vertical Federated Learning (VFL), which draws attention because of its ability to evaluate individuals based on features spread across multiple institutions, encounters numerous privacy and security threats. Existing solutions often suffer from centralized architectures, and exorbitant costs. To mitigate these issues, in this paper, we propose SecureVFL, a decentralized multi-party VFL scheme designed to enhance efficiency and trustworthiness while guaranteeing privacy. SecureVFL uses a permissioned blockchain and introduces a novel consensus algorithm, Proof of Feature Sharing (PoFS), to facilitate decentralized, trustworthy, and high-throughput federated training. SecureVFL introduces a verifiable and lightweight three-party Replicated Secret Sharing (RSS) protocol for feature intersection summation among overlapping users. Furthermore, we propose a (42)-sharing protocol to achieve federated training in a four-party VFL setting. This protocol involves only addition operations and exhibits robustness. SecureVFL not only enables anonymous interactions among participants but also safeguards their real identities, and provides mechanisms to unmask these identities when malicious activities are performed. We illustrate the proposed mechanism through a case study on VFL across four banks. Finally, our theoretical analysis proves the security of SecureVFL. Experiments demonstrated that SecureVFL outperformed existing multi-party VFL privacy-preserving schemes, such as MP-FedXGB, in terms of both overhead and model performance.

The Olera.care Digital Caregiving Assistance Platform for Dementia Caregivers: Preliminary Evaluation Study.

The increasing prevalence of Alzheimer disease and Alzheimer disease-related dementia in the United States has amplified the health care burden and caregiving challenges, especially for caregivers of people living with dementia. A web-based care planning tool, Olera.care, was developed to aid caregivers in managing common challenges associated with dementia care. This study aims to preliminarily evaluate the quality and usability of the Olera.care platform and assess the preferences of using the technology and interests in learning about different older adult care services among caregivers. For interview 1, we aim to understand caregiving needs and let the participants start engaging with the platform. After they engage with the platform, we schedule the second interview and let the participants complete the Mobile Application Rating Scale. The survey also included sociodemographic characteristics, caregiving experiences, communication preferences in technology adoption, and older adult care service use and interests. Descriptive statistics were used to describe the quality and usability of the platform and characteristics of the participants. We conducted 2-sample 2-tailed t tests to examine the differences in the Mobile Application Rating Scale evaluation scores by caregiver characteristics. Overall, 30 adult caregivers in Texas completed the evaluation. The majority were aged ≥50 years (25/30, 83%), women (23/30, 77%), White (25/30, 83%), and financially stable (20/30, 67%). The Olera.care platform evaluation showed high satisfaction, with an overall mean rating of 4.57 (SD 0.57) of 5, and scored well in engagement (mean 4.10, SD 0.61), functionality (mean 4.46, SD 0.44), aesthetics (mean 4.58, SD 0.53), and information quality (mean 4.76, SD 0.44) consistently across all participants. A statistically significant difference (P=.02) was observed in functionality evaluation scores by duration of caregiving, with caregivers dedicating more hours to care rating it higher than those providing less care (mean 4.6, SD 0.4 vs mean 4.2, SD 0.5). In addition, caregivers with less caregiving experience reported significantly higher evaluation scores for aesthetics (P=.04) and information quality (P=.03) compared to those with longer years of caregiving. All participants expressed a willingness to recommend the app to others, and 90% (27/30) rated the app overall positively. Most of the participants (21/30, 70%) favored anonymous interactions before receiving personalized feedback and preferred computer browsers over mobile apps. Medical home health services were the most used, with a diverse range of services being used. Caregiver support groups, medical providers, memory care, meal services, and adult day care were among the most desired services for future exploration. The Olera.care web-based platform is a practical, engaging, easy-to-use, visually appealing, and informative tool for dementia caregivers. Future development and research are essential to enhance the platform and comprehensively evaluate it among a broader population.

Timed-release encryption anonymous interaction protocol based on smart contract

Timed-release encryption (TRE) is a cryptographic primitive that can control the decryption time and has significant application value in time-sensitive scenarios. To solve the reliability issue of nodes in existing TRE anonymous interaction schemes, we propose a blockchain-based TRE protocol for anonymous query time trapdoors. In our protocol, the recipient divides the encrypted trapdoor request information into n ciphertext fragments using secret sharing technology near the decryption time, and employs the idea of onion routing to perform layer-by-layer encryption, creating onion-type data transmitted through middlemen selected from the smart contract. After receiving the ciphertext fragments, the time server integrates them to obtain the trapdoor request information and returns the corresponding time trapdoor to the recipient. This allows the recipient to query any time trapdoor anonymously. Our protocol provides a normative design for the smart contract and specific constraints on the participants’ behavior. Compared with the related anonymous query trapdoor schemes, our protocol improves the probability of successful queries. Security analysis shows that our protocol can resist release-ahead attack, interruption attack, eavesdropping attack, and replacement attack. Performance analysis shows that our protocol outperforms related protocols regarding anonymity, efficiency, and flexibility, achieving highly efficient anonymous interactions. Finally, we conducted an experiment in the Ethereum Rinkeby test network. For the settings of ciphertext fragment number n=3 and ciphertext fragment threshold t=2, the gas consumption for a user to execute the contract was $5.66, which was higher than the contract cost of related schemes, but the contract execution cost was within an acceptable range.

CLHHN: Category-aware Lossless Heterogeneous Hypergraph Neural Network for Session-based Recommendation

In recent years, session-based recommendation (SBR), which seeks to predict the target user’s next click based on anonymous interaction sequences, has drawn increasing interest for its practicality. The key to completing the SBR task is modeling user intent accurately. Due to the popularity of graph neural networks (GNNs), most state-of-the-art (SOTA) SBR approaches attempt to model user intent from the transitions among items in a session with GNNs. Despite their accomplishments, there are still two limitations. First, most existing SBR approaches utilize limited information from short user–item interaction sequences and suffer from the data sparsity problem of session data. Second, most GNN-based SBR approaches describe pairwise relations between items while neglecting complex and high-order data relations. Although some recent studies based on hypergraph neural networks have been proposed to model complex and high-order relations, they usually output unsatisfactory results due to insufficient relation modeling and information loss. To this end, we propose a category-aware lossless heterogeneous hypergraph neural network (CLHHN) in this article to recommend possible items to the target users by leveraging the category of items. More specifically, we convert each category-aware session sequence with repeated user clicks into a lossless heterogeneous hypergraph consisting of item and category nodes as well as three types of hyperedges, each of which can capture specific relations to reflect various user intents. Then, we design an attention-based lossless hypergraph convolutional network to generate sessionwise and multi-granularity intent-aware item representations. Experiments on three real-world datasets indicate that CLHHN can outperform the SOTA models in making a better tradeoff between prediction performance and training efficiency. An ablation study also demonstrates the necessity of CLHHN’s key components.

An Energy Efficiency Based Secure Data Transmission in WBSN Using Novel Id-Based Group Signature Model and SECC Technique

<p>A wireless network composed of wearable sensing along with computing systems connected via a wireless communication channel is termed Wireless Body Sensor Network (WBSN). It enables continuous monitoring through sensors for medical and nonmedical applications. WBSN faces several security problems such as loss of information, access control, and authentication. As WBSN collects vital information and operates in an unfriendly environment, severe security mechanisms are needed in order to prevent the network from anonymous interactions. The different security threats are evaluated with the support of the data transmitted via the sensor networks amongst smart wearable devices. The whole network lifetime together with the Data Transmission (DT) quality is mitigated whilst performing DT utilizing sensor networks, which consume more energy. Hence, in this paper, an energy-efficient secure data transmission mechanism is proposed in WBSN using a novel authentication id-based group signature model and SECC technique. At first, the Group Manager (GM) is selected from the sensors in the remote body sensor system using Normalized Opposition Based Learning BAT Optimization Algorithm (NOBL-BOA). Afterward, clustering with Information Entropy induced K-Means Algorithm (IEKMA) takes place to improve energy efficiency. Next, to provide security to the WBSN, message authentication is carried out based on novel authentication ID-based group signature protocol. Finally, Secret key induced Elliptic Curve Cryptography (SECC) is used to encrypt the message for secure transmission. The simulation results reveal that in comparison with existing works, the proposed work achieves improved security and energy efficiency.</p> <p> </p>

The Formation of the Beginning of Dialogue in Anonymous Online Communication

The main purpose of the article is to track and analyze the correlation of technical and semantic aspects of anonymous interaction on the web. The initial phase of communication between two interlocutors in the “Anonymous VKontakte chat” is empirically investigated. Combining the conceptual and methodological resources of conversational analysis and membership categorization analysis, as well as the theory of affordances adapted to technical objects (in this case, anonymous chat), the author defends the concept of the order of statements of anonymous interlocutors for successful interaction. Most visitors use the chat to search for short-term romantic and/or sexual communication, therefore, at the beginning of the correspondence, the interlocutors often seek to identify each other's gender identity using typical statements. The importance of having the first row is noted, since it is he who, with a high degree of probability, gives the priority right to ask a question about the gender identity of the interlocutor. However, this right does not negate the fact that sooner or later (rather sooner) the owner of the first turn will have to answer the questions of the interlocutor, so the anonymous interaction under consideration, through the communicative efforts of actors organized in the next order, tends to information symmetry.

Effects of sex and sex-related facial traits on trust and trustworthiness: An experimental study.

The ability to trust others, including strangers, is a prerequisite for human cooperation. Economically it is not rational to trust strangers, as trust can be easily exploited. Still, generally, the level of trust toward strangers is relatively high. Trust is closely related to trustworthiness: when trusting others, one expects them to reciprocate. Some individuals elicit more trust than others. Apparently, humans use subtle cues for judging the trustworthiness of their interaction partners. Here, we report on an experiment that investigates trust and trustworthiness in a population of 176 mainly Dutch students. The aims of our study were: (1) to investigate how the sex of interaction partners and their facial appearance (femininity/masculinity) affect the degree of trust and trustworthiness, compared to fully anonymous conditions; (2) to test whether individuals who elicit trust in their interaction partners are trustworthy themselves. Each subject of our experiment played five one-shot Trust Games: one with an anonymous interaction partner, and four "personalized" games after seeing a 20 s silent video of their interaction partner (twice same-sex, and twice opposite-sex). The degree of facial sexual dimorphism was investigated with geometric morphometrics based on full-face photographs. Our results revealed that, despite the already high level of trust in the anonymous setting, the personalization of interactions had a clear effect on behavior. Females elicited more trust in partners of both sexes. Interestingly, females with more feminine faces elicited less trust in both male and female partners, while males with more masculine facial shape were more trusted by females, but less trusted by males. Neither sex nor facial femininity/masculinity predicted trustworthiness. Our results demonstrate that (1) sex and sex-related facial traits of interaction partners have a clear effect on eliciting trust in strangers. However, (2) these cues are not reliable predictors of actual trustworthiness.

Dynamic Min and Max Consensus and Size Estimation of Anonymous Multiagent Networks

In this article, we propose two distributed control protocols for discrete-time multiagent systems, which solve the dynamic consensus problem on the max value. In this problem, each agent is fed an exogenous reference signal and has the objective to estimate and track the instantaneous and time-varying value of the maximum among all the signals fed to the network by exploiting only local and anonymous interactions among the agents. The first protocol achieves bounded steady-state and tracking errors which can be tradedoff for convergence time. The second protocol achieves zero steady-state error and requires knowledge of an upper bound to the diameter of the graph representing the network. Modified versions of both protocols are provided to solve the dual dynamic min-consensus problem. These protocols are then exploited to solve a distributed size estimation problem in a network of anonymous agents in a dynamic setting where the size of the network is time-varying during the execution of the estimation algorithm. Numerical simulations are provided in order to corroborate the characterization of the proposed protocols.

Quality Assurance of the ‘MRCPsych Course’ in Wales

AimsThe ‘MRCPsych Course’ (Membership of the Royal College of Psychiatrists) is provided to all core trainees in psychiatry in Wales by the School of Psychiatry, Health Education and Improvement Wales (HEIW), now delivered online since the start of the COVID-19 pandemic. The aims of the HEIW MRCPsych course are: to prepare core trainees for the MRCPsych exams and to set a ‘robust platform’ for speciality training at the higher level in psychiatry. We undertook a quality assurance of the 2020/21 academic year to see how content and delivery of the course were serving these aims and make recommendations for improvement.MethodsOver the course of one academic year we triangulated trainee feedback, lecturer feedback and peer review. Trainee and trainer feedback forms were sent out following every session. We developed standards and criteria for peer review and reviewed 10% of sessions. We conducted focus groups with trainees using mentimeter to structure a real time, anonymous interaction with parallel verbal and written discussions using a virtual meeting.ResultsTrainee feedback forms were received for 31 lectures from an average of 11 trainees per session. 14 Lecturer feedback forms were received, and 48 trainees attended the two focus groups. 15 hours of teaching underwent peer review.ConclusionLecture content was universally accurate and up to date and all teachers were fluent and engaging, with almost all incorporating research data, guidelines and inspiration for further learning.Several lecturer feedback forms requested a curriculum be provided. Some trainees requested a more exam focused approach with more MCQs.Trainees found online sessions more accessible and convenient. The major downside being that they no longer get to know each other and feel very anonymous, which makes peer support and interaction more difficult.Interactive engagement was the lowest scored domain overall. Interaction seemed to work best when done as a continuous process from the start and when a variety of techniques were engaged. Trainee's suggestions for increasing interactivity included quizzes, polls, breakout rooms, use of interactive tools, and a general encouragement of cameras and microphones on and active discussion throughout the session.Speakers had no problems using the technology to deliver an online session, this triangulated with their high confidence and high satisfaction reported by lectures with HEIW practical support. Trainees reported a high satisfaction with the quality of teaching on the course.Our conclusions have informed changes which are currently being implemented and tested.

Morality and trust in impersonal relationships

This paper investigates the role of moral norms as a social institution in anonymous interactions in trust games. We argue that morality functions as a social norm and therefore comes with specific moral punishments and rewards. Since these moral incentives are confounded with material incentives in the trust game, an augmented version of the game is used, in which moral communication without material consequences is possible. We conduct an experiment that offers participants the opportunity to give morally relevant feedback in ten trust games with random re-matching and under anonymity. In comparison with the original trust game, we find higher levels of both trust and trustworthiness in the augmented game. We also find that these levels remain stable when moral communication is available.

Social Interaction With an Anonymous Opponent Requires Increased Involvement of the Theory of Mind Neural System: An fMRI Study.

An anonymous interaction might facilitate provoking behavior and modify the engagement of theory of mind (TOM) brain mechanisms. However, the effect of anonymity when processing unfair behavior of an opponent remains largely unknown. The current functional magnetic resonance imaging (fMRI) study applied the Taylor aggression paradigm, introducing an anonymous opponent to this task. Thirty-nine healthy right-handed subjects were included in the statistical analysis (13 males/26 females, mean age 24.5 ± 3.6 years). A player winning the reaction-time game could subtract money from the opponent during the task. Participants behaved similarly to both introduced and anonymous opponents. However, when an anonymous opponent (when compared to the introduced opponent) subtracted money, the right inferior frontal gyrus (IFG) demonstrated an increased BOLD signal and increased functional connectivity with the left IFG. Further, increased functional connectivity between the right IFG, the right temporal parietal junction and precuneus was observed during the perception of high provocation (subtracting a large amount of money) from the anonymous compared to the introduced opponent. We speculate that the neural changes may underlie different inferences about the opponents’ mental states. The idea that this reorganization of the TOM network reflects the attempt to understand the opponent by “completing” socially relevant details requires further investigation.

LDGC-SR: Integrating long-range dependencies and global context information for session-based recommendation

Graph neural networks (GNN) have been applied in the session-based recommendation, which aims to predict the potential items the user will interact with next time based on the given anonymous interaction sequences. Most advanced GNN-based recommendation methods (e.g., LESSR, SGNN-HN and GCE-GNN) focus on how to either capture long-range dependencies or use global context information to enhance the recommendation performance. However, these approaches do not consider whether they can enhance each other and cooperatively improve performance. In addition, these methods mainly take the representation of the last interacted item as current interest, ignoring the potential information in other recent interactions and neighbor sessions. To solve these issues, we propose a model called LDGC-SR to integrate Long-range Dependencies and Global Context information for Session-based Recommendation. LDGC-SR employs normalization and adaptive weight fusion mechanism (NAWFM) to integrate long-range node information and global context information to alleviate the over-smooth problem of item representation and the over-fitting problem of the model. NAWFM can reduce the numerical bias between long-range dependencies and global context information, and dynamically adjust their importance to the user’s preference. Moreover, LDGC-SR employs a global context enhanced short-term memory module (GCE-SMM) to capture the user’s current interest more accurately and incorporate it as an auxiliary task. The experimental results demonstrate that our proposed method performs better than the state-of-the-art methods on all four public datasets, and provide suggestions for future session-based recommendation systems.

A novel authentication scheme of identity based group signature algorithm for wireless body sensor network (WBSN)

Wireless Body Sensor Network (WBSN) faces several security problems such as loss of information, access control and authentication. As WBSN collects vital information and operates in unfriendly environment, severe security mechanisms are needed in order to prevent the network from anonymous interactions, authentication is the initial steps towards providing security. An enhanced authentication scheme prevents the system from imposters effectively. In this paper, a novel authentication using identity-based group signature (IBGS) protocol has been proposed to provide security to the WBSN. The proposed method employs identity-based group signature algorithm between biosensors and group manager (GM). An extensive set of experiments were carried out and the results are examined interms of computation cost and energy consumption.

Communication in the Gig Economy: Buying and Selling in Online Freelance Marketplaces

The proliferating gig economy relies on online freelance marketplaces, which support relatively anonymous interactions through text-based messages. Informational asymmetries thus arise that can lead to exchange uncertainties between buyers and freelancers. Conventional marketing thought recommends reducing such uncertainty. However, uncertainty reduction and uncertainty management theories indicate that buyers and freelancers might benefit more from balancing—rather than reducing—uncertainty, such as by strategically adhering to or deviating from common communication principles. With dyadic analyses of calls for bids and bids from a leading online freelance marketplace, this study reveals that buyers attract more bids from freelancers when they provide moderate degrees of task information and concreteness, avoid sharing personal information, and limit the affective intensity of their communication. Freelancers’ bid success and price premiums increase when they mimic the degree of task information and affective intensity exhibited by buyers. However, mimicking a lack of personal information and concreteness reduces freelancers’ success, so freelancers should always be more concrete and offer more personal information than buyers. These contingent perspectives offer insights into buyer–seller communication in two-sided online marketplaces. They clarify that despite, or sometimes due to, communication uncertainty, both sides can achieve success in the online gig economy.

Are facial width-to-height ratio, 2D:4D digit ratio and skeletal muscle mass related to men dominant behavior in the Chicken Game?

Facial width-to-height ratio (fWHR), 2D:4D digit ratio and skeletal muscle mass are morphological traits that have been linked to status-seeking behaviors throughout dominance. However, this link has been contested recently, since the empirical evidence about the relationship between these traits and behavior is mixed. In this study, we tested whether fWHR, 2D:4D digit ratio and skeletal muscle mass were related to dominant behavior employing the Chicken Game, an economic game that may represent a good scenario to investigate hierarchy formation and in which these relationships remain untested. We tested this hypothesis in a sample of 210 Chilean young men (mean = 22.43, SD = 4.35 years old) who played the Chicken Game against an anonymous same-sex individual and one-shot. Our results showed that fWHR was related to dominant choices in the Chicken Game, but null results were found for 2D:4D digit ratio and muscle mass. Accordingly, this study suggests that in a challenging but anonymous interaction only fWHR was related to dominance. Further studies using different conditions of anonymity may contribute to clarify the role of these traits in status-seeking behaviors.

The Role of Culture in Evolutionary Theories of Human Cooperation

Abstract Evolutionarily-minded scholars working on the most puzzling aspects of human cooperation-one-shot, anonymous interactions among non-kin where reputational information is not available-can be roughly divided into two camps. In the first, researchers argue for the existence of evolved capacities for genuinely altruistic human cooperation, and in their models emphasize the role of intergroup competition and selection, as well as group norms and markers of membership that reduce intragroup variability. Researchers in the second camp explain cooperation in terms of individual-level decision-making facilitated by evolved cognitive mechanisms associated with well-established self- and kin-maximization models, as well as by ‘misfires’ that may result from these mechanisms interacting with novel environments. This essay argues that the manner in which culture provides information that de-anonymizes intragroup strangers suggests that neither evolved capacities for genuine altruism nor widespread misfires are necessary to account for anonymous, one-shot cooperation.

Ethnic Markers and How to Find Them : An Ethnographic Investigation of Marker Presence, Recognition, and Social Information.

Ethnic markers are a prominent organizing feature of human society when individuals engage in significant anonymous interactions. However, identifying markers in natural settings is nontrivial. Although ad hoc assignment of markers to groups is widely documented in the ethnographic literature, predicting the membership of individuals based on stylistic variation is less clear. We argue that a more systematic approach is required to satisfy the basic assumptions made in ethnic marker theory. To this end we introduce a three-step ethnographic method to assess the presence, recognition, and transmission of markers of group identity: (1) continual scans, (2) a utilization survey, and (3) a comparative classification task. Applying the method to a study of culturally significant motifs in the South Pacific Island nation of Tonga, we provide evidence that the motif set satisfies basic theoretical assumptions and thus the motifs are likely expressions for social coordination. We also found that the coordinating role of each motif is variable and requires further investigation.

Using targeted betweenness centrality to identify bridges to neglected users in the Twitter conversation on veteran suicide

There are many real-world contexts in which it would be invaluable to identify intermediaries who serve as bridges for specified dyads. For instance, with respect to suicide prevention among military veterans, mental health organizations play an important role in providing information, social support, and other resources to veterans, particularly on social media, where anonymous and pseudonymous interactions may help to counteract the stigma associated with suicide and mental health. However, many at-risk individuals are overlooked in the large-scale conversations on social media and are thus less likely to benefit from those interactions. Intermediaries who can integrate those peripheral actors into the conversation, or at least deliver essential information to them, could help to resolve this issue. The critical question, then, is how to identify those potentially useful intermediaries who are well positioned to act as bridges to reach peripheral actors. To address this challenge, we propose a new network measure, targeted betweenness centrality, to identify vertices that represent potentially useful intermediaries between specified dyads, then use that measure to identify social media users who can act as opinion leaders on behalf of the US Department of Veterans Affairs to reach vertices on the periphery of the Twitter conversation about veteran suicide. The results of this study provide useful insights related to opinion leadership, the military veteran community, and suicide prevention, and more broadly, they demonstrate the practical utility of targeted betweenness centrality for real-world research across a variety of contexts.

Dual Part-pooling Attentive Networks for Session-based Recommendation

Session-based recommendation is proposed to predict user preferences in a short anonymous interaction session. There are two kinds of characteristic in the session: sequential dependencies and collective dependencies. Sequential dependencies mean that user-item interactions in the session are strictly ordered, while collective dependencies mean that multiple interactions with a flexible order jointly determine the user’s next behaviors. However, existing methods only take one of the two characteristics into account, which leads to the fact that they can’t model user behavior well. Furthermore, few methods emphasize the first interaction in the session, although user’s initial intention can be reflected by it to a large extent. To this end, we propose a novel model, Dual Part-pooling Attentive Networks for session-based Recommendation (DPAN4Rec), which is capable of capturing the sequential dependencies and collective dependencies of sessions simultaneously. By proposing a part-pooling attention mechanism, DPAN4Rec explores user’s initial intention from first click and filters out noisy clicks in the session. Extensive experiments have been conducted on two benchmark e-commerce datasets, Yoochoose and Diginetica, and the experimental results show that DPAN4Rec outperforms state-of-the-art methods. Furthermore, our study demonstrates that it is necessary to consider both sequential dependencies and collective dependencies in session-based recommendation.

Characterizing Stage-aware Writing Assistance for Collaborative Document Authoring

Writing is a complex non-linear process that begins with a mental model of intent, and progresses through an outline of ideas, to words on paper (and their subsequent refinement). Despite past research in understanding writing, Web-scale consumer and enterprise collaborative digital writing environments are yet to greatly benefit from intelligent systems that understand the stages of document evolution, providing opportune assistance based on authors' situated actions and context. In this paper, we present three studies that explore temporal stages of document authoring. We first survey information workers at a large technology company about their writing habits and preferences, concluding that writers do in fact conceptually progress through several distinct phases while authoring documents. We also explore, qualitatively, how writing stages are linked to document lifespan. We supplement these qualitative findings with an analysis of the longitudinal user interaction logs of a popular digital writing platform over several million documents. Finally, as a first step towards facilitating an intelligent digital writing assistant, we conduct a preliminary investigation into the utility of user interaction log data for predicting the temporal stage of a document. Our results support the benefit of tools tailored to writing stages, identify primary tasks associated with these stages, and show that it is possible to predict stages from anonymous interaction logs. Together, these results argue for the benefit and feasibility of more tailored digital writing assistance.