# pibully_personalized_cyberbullying_detection_with_peer_influence__47e6c4d5.pdf

PI-Bully: Personalized Cyberbullying Detection with Peer Inﬂuence

Lu Cheng1 , Jundong Li1 , Yasin Silva2 , Deborah Hall3 , Huan Liu1

1Computer Science and Engineering, Arizona State University 2Mathematical and Natural Sciences, Arizona State University 3Social and Behavioral Sciences, Arizona State University {lcheng35,jundongl,ysilva,d.hall,huanliu}@asu.edu

Cyberbullying has become one of the most pressing online risks for adolescents and has raised serious concerns in society. Recent years have witnessed a surge in research aimed at developing principled learning models to detect cyberbullying behaviors. These efforts have primarily focused on building a single generic classiﬁcation model to differentiate bullying content from normal (nonbullying) content among all users. These models treat users equally and overlook idiosyncratic information about users that might facilitate the accurate detection of cyberbullying. In this paper, we propose a personalized cyberbullying detection framework, PI-Bully, that draws on empirical ﬁndings from psychology highlighting unique characteristics of victims and bullies and peer inﬂuence from like-minded users as predictors of cyberbullying behaviors. Our framework is novel in its ability to model peer inﬂuence in a collaborative environment and tailor cyberbullying prediction for each individual user. Extensive experimental evaluations on real-world datasets corroborate the effectiveness of the proposed framework.

1 Introduction

Despite the variability in how cyberbullying is deﬁned [Kowalski et al., 2014] and the extent to which it overlaps with or is viewed as being distinct from cyberaggression [Smith, 2012], there is a general consensus that cyberbullying describes the use of electronic forms of communication to intentionally harm or harass others. Cyberbullying has long been one of the most common online risks for adolescents, however, the rapid growth in the use of social media platforms (e.g., Twitter1) has dramatically increased the potential for cyberbullying to occur. The importance of research that increases the accuracy of cyberbullying detection is underscored by the harmful impact of cyberbullying on victims, which can include negative outcomes such as depression, low self-esteem, and suicidal thoughts and behaviors [Xu et al., 2012]. Within computer science, existing ef-

1https://twitter.com/

forts toward detecting cyberbullying behaviors have primarily focused on building a generic binary classiﬁcation model for all users by analyzing user-generated content [Xu et al., 2012; Dani et al., 2017]. Although these approaches have yielded satisfactory detection performance in practice, these models treat all users equally and, thus fail to capture aspects of cyberbullying that are shaped by the unique characteristics and circumstances of each individual. Empirical ﬁndings within psychology indicate that individual difference variables that is, characteristics that make people different from one another, such as personality traits are important predictors of computer-mediated behaviors [Kowalski et al., 2014; Goodboy and Martin, 2015]. For example, three personality traits, referred to collectively as the Dark Triad, are correlated with cyberbullying perpetration [Goodboy and Martin, 2015]. Speciﬁcally, individuals who exhibit higher levels of (i) machiavellianism (i.e., a desire to manipulate others), (ii) psychopathy (i.e., low levels of empathy), and (iii) narcissism (i.e., feelings of superiority relative to others) are more likely to bully others online [Goodboy and Martin, 2015]. Broadly, these ﬁndings highlight the strong potential for models that characterize and take into account the unique attributes of cyberbullies and their victims to facilitate a better understanding of cyberbullying behaviors. Previous empirical work in psychology has also identiﬁed patterns of similarity in bullying behaviors and victimization within the child and adolescent peer groups, which likely result from peer inﬂuence within established groups, as well as self-selection of youth into friendship groups with similar others [Espelage et al., 2003]. Therefore, given a target user, a key research question is how to model his/her idiosyncratic characteristics and quantify the peer inﬂuence from similar users to facilitate cyberbullying detection. In this paper, we study the novel problem of personalized cyberbullying detection with peer inﬂuence in a collaborative environment. Notably, we use the term personality broadly to refer to users unique collection of traits, characteristics, and circumstances. Building a personalized cyberbullying detection framework that is customized to each individual presents multiple challenges. First, users information in social media platforms is often very noisy, containing irrelevant and redundant features that may jeopardize the learning performance. As a result, it is crucial to building a noise-resilient model to alleviate the negative impact of these uninformative

Proceedings of the Twenty-Eighth International Joint Conference on Artiﬁcial Intelligence (IJCAI-19)

features. Second, in spite of considerable diversity in users personalities, they also share some common attributes and behaviors. In this regard, it is important to capture the commonality shared by all users as well as idiosyncratic aspects of the personality of each individual for automatic cyberbullying detection. Third, in real-world interactions, victims and perpetrators of cyberbullying are inﬂuenced by peers, and the inﬂuence from different users can be quite diverse. Hence, developing a way to encode the diversity of peer inﬂuence for cyberbullying detection is imperative. The main contributions of this paper are as follows:

We formally deﬁne the problem of personalized cyberbullying detection with peer inﬂuence in a collaborative environment. The core idea of our formulation is to customize the prediction for individuals.

We propose a novel cyberbullying detection framework which consists of three components: (1) a global component that identiﬁes the commonalities among all users; (2) a personalized component that captures the idiosyncratic characteristics of each individual; and (3) a collaborative/peer inﬂuence component that can quantify the diverse inﬂuence from other users.

We perform empirical experiments on multiple realworld datasets from microblogging platforms to corroborate the efﬁcacy of the proposed framework.

The remainder of this paper is organized as follows. In Section 2, we formally deﬁne the problem of personalized cyberbullying detection with peer inﬂuence in a collaborative environment. Section 3 describes the proposed PI-Bully framework in detail. In Section 4, we discuss the used datasets, the experimental settings, and our ﬁndings from experiments. Section 6 reviews related work and Section 7 concludes the paper and describes some paths for future work.

2 Problem Statement

Suppose U users generate N posts in a social media platform. Let {(xi j, yi j) | j = 1, ..., Ni} represent the posts from the i-th user and deﬁne N = PU i=1 Ni. Each speciﬁc post j from user i is represented by (xi j, yi j). xi j RD represents the post s features, D is the number of features, and yi j denotes the class label associated with the post. In this work, we assume each post is associated with two possible labels yi j {0, 1}, where yi j = 1 denotes that the post is a cyberbullying message and yi j = 0 otherwise. Then X = [(x1 1)T ; ...; (x1 N1)T ; ...; (x U 1 )T ; ...; (x U NU )T ] RN D is the feature representation of all these N posts and y = [y1 1; ...; y1 N1; ...; y U 1 ; ...y U NU ] {0, 1}N is the corresponding label vector. With the aforementioned notations, we deﬁne the problem of personalized cyberbullying detection with peer inﬂuence in a collaborative environment as follows: Given the feature representation X of N social media posts from U users and the label vector y of these N posts, the goal is to train a binary classiﬁcation model to predict the labels of online social media posts (bullying or normal). In particular, during the learning phase, we would like to (1) tailor the prediction for each user by capturing commonalities

among multiple users and individual characteristics; and (2) quantify the way each user is inﬂuenced by like-minded users.

3 The Proposed Framework

In this section, we describe how to build a generic classiﬁcation model to identify cyberbullying behaviors. We ﬁrst show a global model that is designed to capture the commonality shared by all users and then describe the mechanisms to model users idiosyncrasies. In addition, as the occurrence of cyberbullying is heavily related to peer inﬂuence, we investigate how to quantify the inﬂuence from like-minded users such that personalized modeling can beneﬁt from users with similar behaviors. Finally, we show how to predict an unlabeled post from an unseen user using the PI-Bully model and brieﬂy introduce the optimization algorithm and its time complexity. Fig. 1 illustrates the overview workﬂow of the PI-Bully framework.

3.1 Building the Personalized Model

Previous efforts in cyberbullying detection have been primarily devoted to the development of a global classiﬁcation model to capture the commonalities among users. It formulates cyberbullying detection as a binary classiﬁcation task:

j=1 f(xi j, yi j, w) + λ1 w 1, (1)

where w RD is the global classiﬁcation model that applies to all users. As vast majority of feature representation methods for social media posts may lead to the inclusion of uninformative features, we integrate feature selection [Li et al., 2017] into the classiﬁer by imposing an ℓ1norm sparse regularization term, where λ1 controls the sparsity of the model. f( ) is a loss function to measure the loss between the ground truth class labels and predicted class labels. In this work, we use the squared loss function, i.e., f(xi j, yi j, w) = (w T xi j yi j)2, but the model could also use other functions such as hinge loss and cross entropy loss. In spite of the empirical success of global classiﬁcation models, research advances in psychology indicate that cyberbullying is correlated with a number of individual features such as personality traits (e.g., [Baughman et al., 2012]), attitudes and beliefs (e.g., [Hinduja and Patchin, 2013]), and motives (e.g., [Gradinger et al., 2011]) that vary from user to user. In short, although users may share a number of inherent characteristics, they are also highly idiosyncratic. To this end, we assume each user ui has a personalized model Mi RD in addition to the global model w RD. Moreover, we impose an ℓ1-norm sparse regularization term on each personalized model Mi to reduce the model complexity. Hence, we obtain the following optimization framework:

j=1 f(xi j, yi j, w + Mi) + λ1( w 1 +

Proceedings of the Twenty-Eighth International Joint Conference on Artiﬁcial Intelligence (IJCAI-19)

0.2 0.2 0.6

Uses shared and personalized features

Test data Predicted label

For unknown users, measures the influence of known users on the new user and estimates model parameters

For known users, uses already learned model parameters

Figure 1: The proposed PI-Bully framework. We ﬁrst leverage the data matrix X to compute the inﬂuence matrix S, which quantiﬁes how users could inﬂuence each other. For example, u2 is inﬂuenced more by u1 than u4 as s21 = 0.9 is larger than s24 = 0.1. Then, in the training phase, we use both shared and user-speciﬁc features to train a classiﬁer by capturing the commonalities of all users and the idiosyncrasies of each speciﬁc user. Finally, in the testing phase, given a set of unlabeled test data, we predict if the new posts are cyberbullying or not.

3.2 Characterizing Peer Inﬂuence The process of model parameter learning in the above personalized model can be problematic due to the limited amount of training data for each user. Because of this, the generated personalized model can easily suffer from overﬁtting and have poor generalization ability on unseen test data. To address this problem, we decompose the personalized model Mi of each user into a personalized component, Pi RD, which encodes a user s inherent traits, and a collaborative/peer inﬂuence component, Qi RD. The goal for including this collaborative/peer inﬂuence component is to extrapolate information about the way a user experiences cyberbullying from the experiences of similar users. By doing this, we aim to capture the inﬂuence of similar, like-minded users in the way a person experiences cyberbullying. The collaborative/peer inﬂuence component Qi is customized for each user and is estimated by a weighted average of the personalized component Pi of other users. The integration of this component is motivated by empirical ﬁndings in psychology indicating similarity within child and adolescent peer groups in both bullying behaviors and victimization [Espelage et al., 2003]. Then the objective function in Eq. (2) can be reformulated as

min w,Pi,Qi

j=1 f(xi j, yi j, w + Pi + Qi)+

λ1( w 1 + P 1) + λ2

j=1 sji Pj 2 2,

where P, Q RU D respectively denote the concatenation of personalized component and collaborative component of all users, λ2 balances the contribution of collaborative/peer inﬂuence for personalized cyberbullying detection and sji S denotes how user ui is inﬂuenced by user uj (the inﬂuence from different peers could vary signiﬁcantly). Here, we provide a more intuitive illustration of these com-

ponents. As the proposed model focuses mainly on text data, w captures the common language used globally across all the users, P represents the unique language characteristics of individual users, and Q is a model parameter that captures additional predictive value that can be drawn from between-user language similarities. In this work, we exploit the method presented in [Anava and Levy, 2016] to adaptively learn the optimal neighborhood structure (i.e., the most inﬂuential neighbors) around each user to quantify the diversity of peer inﬂuence. Speciﬁcally, we leverage the k -NN algorithm in [Anava and Levy, 2016] to process the data matrix X and generate the User-to User peer inﬂuence matrix S of size U U. The i-th row in S represents the pairwise similarities between user i and other users. In Eq. (3), we can also observe that the personalized models Pi of similar users are explicitly correlated with each other through the last term (collaborative/peer inﬂuence component), which implicitly generates additional data for each user to train the personalized model. In summary, we can observe that the PI-Bully model for each user ui has three components: (1) a global model w that captures the shared traits of all users; (2) a personalized model Pi that captures the unique characteristics of the user; and (3) a collaborative/peer inﬂuence component that quantiﬁes how the user is inﬂuenced by like-minded users.

3.3 Inference on Unlabeled Data

Next, we describe how the PI-Bully framework predicts whether an unlabeled post x RD is a cyberbullying message or not, given the learned parameters w, P, Q. There are two cases to discuss. If the user u of the post x appears in the training dataset, we can directly use the learned model parameters to make the prediction. In this case, the classiﬁer for the new post x is c = w + Pu + Qu. If the post is from a new user m that does not appear in the training dataset, we ﬁrst leverage the same mechanism as the one used

Proceedings of the Twenty-Eighth International Joint Conference on Artiﬁcial Intelligence (IJCAI-19)

in [Anava and Levy, 2016] to measure the inﬂuence of existing users in the training dataset on user m. Then, we estimate the model parameters for this new user by solving the following optimization problem (a.k.a. Weber problem [Hallac et al., 2015]):

i=1 sim( Pm Pi 2 + Qm Qi 2), (4)

where sim indicates how the new user m is inﬂuenced by user i. After solving the above Weber problem, the classiﬁer for the new post is speciﬁed as c = w + Pm + Qm. Using the classiﬁer c for the new unlabeled post, we predict that the new post is cyberbullying if x T c 0.5, and normal otherwise.

3.4 Optimization The proposed PI-Bully model in Eq. (3) has three sets of model parameters: w, Pi (i = 1, ..., U) and Qi (i = 1, ..., U). We can observe that the objective function is not convex regarding these three sets of parameters simultaneously. In addition, it is not smooth as well due to the ℓ1norm sparse regularization terms. Motivated by [Wu and Huang, 2016], we address these problems by using the Alternating Direction Method of Multipliers (ADMM) [Boyd et al., 2011] and Fast Iterative Shrinkage-Thresholding (FISTA) [Beck and Teboulle, 2009] to achieve a local optimal solution. Details are omitted here due to the space constraint.

4 Experimental Evaluation In this section, we present experimental results to evaluate the effectiveness of the proposed PI-Bully model. In particular, we aim to answer the following research questions: (1) Can the proposed framework achieve better cyberbullying detection performance than existing models? (2) What is the impact of the different components of the proposed PI-Bully framework? and (3) How robust is the proposed model w.r.t. different model hyperparameters?

4.1 Datasets We use two real-world datasets crawled from the microblogging platform, Twitter2. The ﬁrst dataset3 (referred to Xu et.al), with 3,095 social posts, was published in [Xu et al., 2012]. Note that the original dataset consists of 7,321 tweets, among which only 3,095 tweets were publicly available at the time we crawled using the Tweet IDs. Following the procedure suggested by [Nand et al., 2016], we collect the second dataset (referred as Authors) via the Twitter streaming API from September 19th to 25th, 2017 using the following keywords: nerd, gay, loser, freak, emo, whale, pig, fat, wannabe, poser, whore, should, die, slept, caught, suck, slut, live, afraid, ﬁght, pussy, cunt, kill, dick, bitch. Early in the experimental design phase, a decision was made to maximize the labeling quality (while maintaining a proper dataset size) instead of prioritizing only the data size. Following this guideline, we extracted 20,000 tweets to be manually labeled by well-trained human annotators (with

2https://twitter.com/ 3http://research.cs.wisc.edu/bullying/data.html

Datasets # Users # Tweets # Bullying # Normal Xu et al. 2,948 3,095 1,794 1,301 Authors 9,833 19,994 3,845 16,149 Table 1: Basic statistics of the two used datasets.

backgrounds in psychology and computer science) over a period of two months. The human annotators followed coding guidelines that were similar to the ones described in [Nand et al., 2016]. Each tweet was initially labeled by two annotators and the agreement level between the two annotators in this stage was 80%. A third annotator was asked to resolve the conﬂicts identiﬁed in the initial annotation phase. After conﬂict resolution and data cleaning, we ﬁnally obtained the Authors dataset with a total number of 19,994 labeled tweets. Table 1 shows the statistics for these two datasets. It is important to note that the proportions of bullying to normal messages in these two datasets are different. Whereas 57.96% of the posts in the Xu et al. dataset are bullying messages, 19.23% of the posts in the Authors dataset are bullying interactions. This latter percentage is similar to the one found in [Hosseinmardi et al., 2015] for Instagram data and more closely represents the proportion of bullying to non-bullying messages in the real-world. Our dataset can be downloaded from http://www.public.asu.edu/ lcheng35/. We perform psychometric analysis to obtain features for each tweet in the aforementioned datasets through Linguistic Inquiry Word Count (LIWC) [Pennebaker et al., 2001]. Speciﬁcally, LIWC counts words that belong to certain categories in psychology. For example, the word cry belongs to ﬁve categories: sadness, negative emotion, overall affect, verb, and past tense verb. The results of previous research show that such psychometric analysis can improve the performance of cyberbullying detection [Nand et al., 2016].

4.2 Performance Evaluation To answer the ﬁrst question, we compare PI-Bully with common text classiﬁcation models (k NN, Random Forest, Linear SVM, and Logistic Regression) with the same input features and two text-based cyberbullying detection models (Bully [Xu et al., 2012] and SICD [Dani et al., 2017]). We specify these models below.

k NN: It predicts the class labels of unlabeled instances using a k-nearest classiﬁer where the distance metric is speciﬁed as the Euclidean distance.

Random Forest (RF): It is an ensemble learning method that constructs a multitude of decision trees during training and output the mode of the classes at testing.

Linear SVM (SVM): It implements a regularized linear support vector machine model with stochastic gradient descent (SGD) learning.

Logistic Regression (LR): It is an extension of linear regression model for the classiﬁcation problem with the logistic function as the loss function.

Bully [Xu et al., 2012]: This model extracts several NLP features including unigrams, unigrams+bigrams, and POS colored N-grams to train a SVM model.

Proceedings of the Twenty-Eighth International Joint Conference on Artiﬁcial Intelligence (IJCAI-19)

Metrics Precision Recall F1 AUC k NN 0.663 0.364 0.470 0.652 SVM 0.699 0.469 0.562 0.701 RF 0.708 0.478 0.571 0.707 LR 0.680 0.485 0.566 0.705 Bully 0.653 0.508 0.571 0.709 SICD 0.803 0.263 0.396 0.791 PI-Bully 0.425 0.887 0.574 0.844 Table 2: Performance comparison w.r.t. Authors dataset.

Metrics Precision Recall F1 AUC k NN 0.663 0.517 0.581 0.662 SVM 0.685 0.646 0.665 0.714 RF 0.681 0.544 0.605 0.678 LR 0.680 0.646 0.663 0.711 Bully 0.708 0.646 0.676 0.725 SICD 0.727 0.609 0.663 0.722 PI-Bully 0.656 0.740 0.695 0.802 Table 3: Performance comparison w.r.t. Xu et al. dataset.

SICD [Dani et al., 2017]: It uses both content (i.e., TFIDF) and sentiment information embedded in the usergenerated content to boost the performance of cyberbullying detection.

Our evaluation methods include several widely-used metrics - Precision, Recall, F1 score, and AUC. The main reason we choose F1 score and AUC rather than Accuracy is that the cyberbullying datasets are typically imbalanced, i.e., each class does not make up an equal proportion of the dataset. Meanwhile, imbalanced datasets may affect the trade-off between recall and precision. In the context of cyberbullying detection, missing a positive instance is usually less desirable than incorrectly labeling a negative instance. Hence, achieving high recall is particularly important. In the experiments, we use 80% of the datasets for training and the rest for testing, the averaged classiﬁcation results based on ten runs are shown in Tables 2-3. We select the hyperparameters based on cross-validation on the training data. A detailed hyperparameter analysis can be seen in Sec. 4.4. The best and the second best scores are highlighted with bold and underscored text, respectively. We can observe that for both datasets, PI-Bully achieves the best Recall, F1, and AUC scores while most baseline models present poor Recall and AUC scores. This is especially apparent for Authors dataset, which more closely represents the proportion of bullying to normal messages in the real-world. We can conclude that PI-Bully signiﬁcantly boosts the classiﬁcation of positive samples and leads to the improved overall performance of cyberbullying detection. Results of a pairwise Wilcoxon signed-rank test indicate that the improvement of PI-Bully is signiﬁcant, with a 0.05 signiﬁcance level.

4.3 Impact of Different Model Components

For each user, PI-Bully is composed of three components: (1) the global model w common to all users; (2) a personalized component Pi that is customized for each individual user;

20% 40% 60% 80% Percentages of training data

P G G+P G+I PI-Bully

Figure 2: Performance evaluation of different components on the Authors dataset.

and (3) a collaborative/peer inﬂuence component that quantiﬁes the inﬂuence from like-minded users. We compare in this subsection the following variants: The personalized component (P): a variant of the proposed PI-Bully framework that only includes the personalized component Pi for each user. The global model (G): a variant that only includes the global model w. Global+Personalized (G+P): a variant of PI-Bully without the peer inﬂuence component. Global+Inﬂuence (G+I): a variant of PI-Bully that eliminates the personalized component Pi for each user. We compare these four variants with the proposed PI-Bully model using the Authors dataset. For these experiments, the percentage of training data is incrementally increased from 20% to 80%. The comparison of these four variants and the proposed PI-Bully model is illustrated in Fig. 2. We highlight the following key ﬁndings: The personalized component P is inferior to the global model G. The main reason is that P often suffers from the over-parameterization issue due to the lack of training data, whereas the global model G can collect more data to capture the commonalities among all users during the training phase. Both the G+I model and the G+P model outperform the global model G. These results validate the importance of incorporating the personalized and peer inﬂuence components for personalized cyberbullying detection. The proposed PI-Bully framework achieves the best performance and the dominance tends to become more obvious as the dataset becomes more imbalanced, i.e. the Authors dataset. This shows the beneﬁts of considering the three proposed components.

4.4 Hyperparameter Analysis The PI-Bully model has two key hyperparameters: λ1 controls the complexity of the model (i.e., it balances the sparsity of personalized features and common features in the model learning phase) while λ2 regulates the importance of peer inﬂuence in PI-Bully. To investigate the effect of these two

Proceedings of the Twenty-Eighth International Joint Conference on Artiﬁcial Intelligence (IJCAI-19)

1e-7 1e-5 1e-3 0.1 10 Values

Figure 3: Effects of λ1, λ2 w.r.t. AUC on the Authors dataset.

parameters, we ﬁx one parameter at a time (λ1=1e-7,λ2=1e-7 respectively) and vary the other one to evaluate how it affects the classiﬁcation performance. We vary the values of λ1 and λ2 among {1e-7,1e-5,1e-3,0.1,10} and show the AUC score in Fig. 3. We can see the performance keeps stable when λ1 is speciﬁed as a small number, while large λ1 overemphasizes the personalized features can result in relatively poor performance. Consequently, a proper value of λ1 enables the identiﬁcation of the most representative personalized features for both positive and negative samples. The proposed framework is more robust to changes of λ2, and presents an increasing trend when λ2 is in a certain range. In summary, the performance of PI-Bully is relatively stable when the hyperparameters are varied in a certain range, and thus can be tuned for various application purposes.

5 Related Work

Cyberbullying is a serious issue with signiﬁcant negative societal consequences. To date, a number of dedicated learning algorithms have been proposed to identify cyberbullying instances. Most existing methods adopt a two-stage approach to detect cyberbullying: they ﬁrst apply feature engineering to identifying feature sets that enable capturing cyberbullying patterns and then employ off-the-shelf machine learning classiﬁers to detect cyberbullying behaviors. Typically, the feature set includes text-based [Xu et al., 2012; Dani et al., 2017; Bellmore et al., 2015] and network-based features [Squicciarini et al., 2015; Al-garadi et al., 2016]. Various methods differ in the types of features used for classiﬁcation. For example, Dinakar et al. [Dinakar et al., 2011] concatenated TF-IDF features, POS tags of frequent bigrams, and profane words as content features to detect cyberbullying behaviors. Xu et al. [Xu et al., 2012] presented several offthe-shelf tools such as Bag-of-Words models and LSAand LDA-based representation learning to predict bullying traces in Twitter. In [Dadvar et al., 2013], the authors made use of gender-speciﬁc features and contextual features, such as users previous posts and the use of profane words, to improve the performance of cyberbullying detection. Dani et al. [Dani et al., 2017] proposed the SICD model which incorporates sentiment into content features. Their goal was to facilitate cyberbullying detection by capturing the sentiment consistency of normal and bullying posts. Bellmore

et al. [Bellmore et al., 2015] used a dictionary including words in a Twitter corpus to construct a frequency vector for each tweet and trained a text classiﬁer to answer core questions about cyberbullying ( Who, What, Why, Where, and When ). With the increasing prevalence of social networking systems, network-based features (e.g., the number of friends, network embeddedness, and relational centrality) are also used to detect cyberbullying behaviors [Squicciarini et al., 2015]. For instance, previous work by [Al-garadi et al., 2016] studied a model that integrated the use of activity information, user information, and tweet content features. Cyberbullying has also been studied in other social media platforms such as Ask.fm [Li et al., 2014], Instagram [Hosseinmardi et al., 2015; Cheng et al., 2019], and Vine [Raﬁq et al., 2016]. In addition, some work has focused on developing systems and applications to identify cyberbullying behaviors on social network platforms [Silva et al., 2016a; 2016b]. The authors aim to estimate the probability of an individual experiencing cyberbullying considering the received messages and various cyberbullying risk factors.

6 Conclusions and Future Work

Existing efforts toward detecting cyberbullying have heavily focused on building generic classiﬁcation models for all users that seek to distinguish bullying behaviors from normal content. These methods, however, ignore unique characteristics that are embedded in the user-generated content. Empirical ﬁndings from psychology highlight the role of individual difference variables reﬂected in users unique personality traits, attitudes, motives, etc. and inﬂuence from like-minded users as predictors of cyberbullying. In this paper, we propose a principled personalized cyberbullying detection framework, PI-Bully, that draws on these interdisciplinary ﬁndings to tailor and improve the prediction of cyberbullying behaviors. Future work in cyberbullying detection can be performed in several key areas. First, there has been limited research examining predictive models that take temporal properties and patterns of cyberbullying into account, which stands to contribute to a deeper understanding of the nature of cyberbullying across research disciplines. Second, there is a growing need for cyberbullying detection models that rely on limited or aggregated data in part, due to the difﬁculty of accessing social media data. This underscores the need for models that can achieve high accuracy while relying on limited, incomplete, anonymized, or aggregated data.

Acknowledgements

This material is based upon work supported by the National Science Foundation (NSF) Grant 1719722.

References [Al-garadi et al., 2016] Mohammed Ali Al-garadi, Kasturi Dewi Varathan, and Sri Devi Ravana. Cybercrime detection in online communications: The experimental case of cyberbullying detection in the twitter network. Computers in Human Behavior, 63:433 443, 2016.

Proceedings of the Twenty-Eighth International Joint Conference on Artiﬁcial Intelligence (IJCAI-19)

[Anava and Levy, 2016] Oren Anava and Kﬁr Levy. k*- nearest neighbors: From global to local. In NIPS, pages 4916 4924, 2016. [Baughman et al., 2012] Holly M Baughman, Sylvia Dearing, Erica Giammarco, and Philip A Vernon. Relationships between bullying behaviours and the dark triad: A study with adults. Personality and Individual Differences, 52(5):571 575, 2012. [Beck and Teboulle, 2009] Amir Beck and Marc Teboulle. A fast iterative shrinkage-thresholding algorithm for linear inverse problems. SIIMS, 2(1):183 202, 2009. [Bellmore et al., 2015] Amy Bellmore, Angela J Calvin, Jun-Ming Xu, and Xiaojin Zhu. The ﬁve w s of bullying on twitter: who, what, why, where, and when. Computers in human behavior, 44:305 314, 2015. [Boyd et al., 2011] Stephen Boyd, Neal Parikh, Eric Chu, Borja Peleato, and Jonathan Eckstein. Distributed optimization and statistical learning via the alternating direction method of multipliers. Foundations and Trends R in Machine Learning, 3(1):1 122, 2011. [Cheng et al., 2019] Lu Cheng, Jundong Li, Yasin Silva, Deborah Hall, and Huan Liu. Xbully: Cyberbullying detection within a multi-modal context. In WSDM, pages 339 347, 2019. [Dadvar et al., 2013] Maral Dadvar, Dolf Trieschnigg, Roeland Ordelman, and Franciska de Jong. Improving cyberbullying detection with user context. In ECIR, pages 693 696, 2013. [Dani et al., 2017] Harsh Dani, Jundong Li, and Huan Liu. Sentiment informed cyberbullying detection in social media. In ECML PKDD, pages 52 67, 2017. [Dinakar et al., 2011] Karthik Dinakar, Roi Reichart, and Henry Lieberman. Modeling the detection of textual cyberbullying. The Social Mobile Web, 11(02), 2011. [Espelage et al., 2003] Dorothy L Espelage, Melissa K Holt, and Rachael R Henkel. Examination of peer group contextual effects on aggression during early adolescence. Child development, 74(1):205 220, 2003. [Goodboy and Martin, 2015] Alan K Goodboy and Matthew M Martin. The personality proﬁle of a cyberbully: Examining the dark triad. Computers in human behavior, 49:1 4, 2015. [Gradinger et al., 2011] Petra Gradinger, Dagmar Strohmeier, and Christiane Spiel. Motives for bullying others in cyberspace. Cyberbullying in the global playground: Research from international perspectives, 263, 2011. [Hallac et al., 2015] David Hallac, Jure Leskovec, and Stephen Boyd. Network lasso: Clustering and optimization in large graphs. In KDD, pages 387 396, 2015. [Hinduja and Patchin, 2013] Sameer Hinduja and Justin W Patchin. Social inﬂuences on cyberbullying behaviors among middle and high school students. Journal of youth and adolescence, 42(5):711 722, 2013.

[Hosseinmardi et al., 2015] Homa Hosseinmardi, Sabrina Arredondo Mattson, Rahat Ibn Raﬁq, Richard Han, Qin Lv, and Shivakant Mishra. Analyzing labeled cyberbullying incidents on the instagram social network. In Soc Info, pages 49 66, 2015. [Kowalski et al., 2014] Robin M Kowalski, Gary W Giumetti, Amber N Schroeder, and Micah R Lattanner. Bullying in the digital age: A critical review and meta-analysis of cyberbullying research among youth., 2014. [Li et al., 2014] Homa Hosseinmardi Shaosong Li, Zhili Yang, Qin Lv, Rahat Ibn Raﬁq Richard Han, and Shivakant Mishra. A comparison of common users across instagram and ask. fm to better understand cyberbullying. In Bd Cloud, pages 355 362, 2014. [Li et al., 2017] Jundong Li, Kewei Cheng, Suhang Wang, Fred Morstatter, Robert P Trevino, Jiliang Tang, and Huan Liu. Feature selection: A data perspective. CSUR, 50(6):94, 2017. [Nand et al., 2016] Parma Nand, Rivindu Perera, and Abhijeet Kasture. how bullying is this message? : A psychometric thermometer for bullying. In COLING, pages 695 706, 2016. [Pennebaker et al., 2001] James W Pennebaker, Martha E Francis, and Roger J Booth. Linguistic inquiry and word count: Liwc 2001. Mahway: Lawrence Erlbaum Associates, 71(2001):2001, 2001. [Raﬁq et al., 2016] Rahat Ibn Raﬁq, Homa Hosseinmardi, Sabrina Arredondo Mattson, Richard Han, Qin Lv, and Shivakant Mishra. Analysis and detection of labeled cyberbullying instances in vine, a video-based social network. Social Network Analysis and Mining, 6(1):88, 2016. [Silva et al., 2016a] Yasin N. Silva, Christopher Rich, Jaime Chon, and Lisa M. Tsosie. Bullyblocker: An app to identify cyberbullying in facebook. In ASONAM 2016,, pages 1401 1405, 2016. [Silva et al., 2016b] Yasin N. Silva, Christopher Rich, and Deborah Hall. Bullyblocker: Towards the identiﬁcation of cyberbullying in social networking sites. In ASONAM 2016, pages 1377 1379, 2016. [Smith, 2012] Peter K Smith. Cyberbullying and cyber aggression. In Handbook of school violence and school safety, pages 111 121. Routledge, 2012. [Squicciarini et al., 2015] A Squicciarini, S Rajtmajer, Y Liu, and Christopher Grifﬁn. Identiﬁcation and characterization of cyberbullying dynamics in an online social network. In ASONAM, pages 280 285, 2015. [Wu and Huang, 2016] Fangzhao Wu and Yongfeng Huang. Personalized microblog sentiment classiﬁcation via multitask learning. In AAAI, pages 3059 3065, 2016. [Xu et al., 2012] Jun-Ming Xu, Kwang-Sung Jun, Xiaojin Zhu, and Amy Bellmore. Learning from bullying traces in social media. In NAACL HLT, pages 656 666, 2012.

Proceedings of the Twenty-Eighth International Joint Conference on Artiﬁcial Intelligence (IJCAI-19)