Please support the Artimys Spitball & Fireball Kickstarter campaign!! Click here
The Most Accurate
Means No Spying
A Social Solution
to a Social Problem
See at a glance how active your child has been online since the last report. See whom your child has been speaking with most. And track their overall activity to look out for significant changes in online activity. See how our threat analyzers score your child's conversations. Look out for conversations that that are flagged by our system as likely threats as well as look out for so called "outliers"; conversations that are scored significantly higher than your child's other conversations.
Artimys gives a comprehensive view of your child's messages at a glance. Each message thread is scored based on the likelihood it contains a sexual predator, bullying or suicide risk. Artimys then arranges these scores according to how many times each score occurs. For example, if 5 of your child's message threads got a score of 67 for sexual predator threat, the bar over 67 will be 5 units tall. This allows you to glance at the chart and quickly determine if any scores are high enough to require investigation. In the example to the right, two threads got a score of 95 for sexual predators, which is cause for concern. Also, if a small set of threads have an abnormally high score relative to the others, this indicates those conversations may require investigation.
Compared with all other security products that only flag certain words, Artimys reads full messages and threads and flags potential threats. The result is a clear view of the threats facing your child and a better ability to act on them.
Artimys vs Leading Family Security Products
An experiment was performed in which two threads were presented to Artimys as well as a leading family security product. One of the threads is teen chat from a popular online forum and the other thread is an actual transcript of a predator seducing a child.
Artimys rated the predator chat a score of 93 and the teen chat a score of 46. The 93 score is high and would be an immediate red flag alert to a parent. The leading security product identified the teen chat as dangerous because the word bullshit was used, but did not flag the predator chat. Artimys considered the entire message thread and found the tell-tale pattern of a sexual predator while ignoring the foul, but not dangerous language. The leading security product suffered from a false-positive and a false-negative failure. Only Artimys considers the entire context of electronic messages and identifies the language and emotional patterns of threats. No other service comes close to the accuracy and flexibility of Artimys.
Pedophiles are mentally ill. They are also incredibly smart, driven and determined experts whose predatory skills are constantly evolving and improving. Even the most attentive and well-intentioned parents are amateurs who have neither the expertise nor the time to guard their children against these skilled predators.
Possessing the skill and experience of trained human experts and professionals; Artimys is the first fully automated predator identification tool that truly helps parents protect their children. By analyzing hundreds of examples of actual predators in acts of online seduction, we have systematically trained computers to identify the unique language and emotional “fingerprint” of pedophiles in action.
From a recent study conducted for the US Sentencing Commission:
“Comparison studies have confirmed that online sexual offenders differ from conventional sexual offenders in meaningful ways, including differences in average age, education, prior criminal history, and the psychological risk factors of sexual deviance, sexual preoccupation, and sexual self-regulation (Babchishin et al., 2011; Seto, Wood et al., in press). Online offenders are younger, on average, better educated, and have less prior criminal history than contact offenders. Online offenders scored higher on sex-related psychological risk factors, but score lower on criminal risk factors such as criminal history and antisocial personality traits, suggesting that online offenders may pose a lower risk to reoffend overall than contact offenders.” †
Pedophiles are born as pedophiles and their condition has been found to be untreatable. ††
†: Michael C. Seto, P. (2012). Child Pornography Offender Characteristics and Risk to Reoffend. Ottawa: Prepared for the United States Sentencing Commission.
††: Gilbert, H. (2010). Pessimism about pedophilia. Harvard Mental Health Letter , 1-3.
Bullying is not new, it's been around as long as there have been children. What is new is the use of the Internet for bullying. The Internet allows bullies to easily target specific children and readily access them. Bullies can remain annonymous and/or misrepresent their identity. This not only increases how much they bully, but also increases how viciously they bully. All of the tools that are used by business to make their operations more efficient and effective are being used by bullies to make their bullying more efficient and more effective. The Internet didn't create bullying, but it gave the bully even greater advantage over their victims.
Research conducted by the Center for Disease Control shows that the majority of bullying occurs in private messages. 67% of children report being bullied via Instant Messaging compared with 25% that report being bullied in a chat room. The effects of bullying are profound. They range from feelings of anxiety to withdrawal to extreme cases of suicide.
Artimys protects your child against bullies by learning to identify the language of bullies. Artimys has employed hundreds of people to examine millions of electronic messages. These messages have been classified as either containing bullying or not. Artimys then uses these messages to learn the patterns that distinguish bullying from healthy communications. By examining the full context of the message or the thread that the message is in, and not just specific words, Artimys is able to detect a wide range of bullying techniques while distinguishing this from playful interaction.
Teen suicide is the third leading cause of death among people aged 15 to 24 behind homocide and accidents (U.S. Center for Disease Control and Prevention). But teen suicide is preventable. Research has shown that suicidal children show signs of depression or emotional distress. Artimys protects your children from suicide risk by detecting suicidal and morbid discussion in your child's electronic communications. Artimys has assembled a massive database of actual teens discussing suicide and expressing feelings indicative of depression. Using these real-world communications, Artimys learns to identify the patterns of depression and morbidity. And by examining the full context of the discussion and your child's other conversations (compared to looking just for specific words), Artimys can provide you with an accurate assesment of your child's suicide risk. No other service can compare to the Artimys Guradian's level of protection.
Artimys uses Text Classification to distinguish message threads that contain threats from those that do not.
Text classification is a technology used to sort documents based on their content. Artimys uses Text Classification technology to identify threats in messages. For instance, search engines use Text Classification to retrieve documents and web pages. Spam filters use Text Classification technology to distinguish email messages that are junk mail from non junk mail.
Artimys combines sophisticated psychometric analysis with linguistic analysis. The result is a powerful system that exceeds the abilities of human experts both in terms of it’s accuracy and it’s vigilance. And because online threats exist online, the Artimys Guardian is constantly running searches for new examples of these threats and incorporates them into the portfolio of patterns (“fingerprints”) that are detected. The system adapts to provide up to date protection as the threats adapt and change.
Text Classification has been studied extensively and shown to be as good or better than human experts at classifying text:
“The accuracy of modern text classification systems rivals that of trained human professionals, thanks to a combination of information retrieval (IR) technology and machine learning (ML) technology.” (Sebastiani, 2005)
For the task of classifying texts in general, computers have been shown to have an accuracy that’s greater than 90% (Mehran, Yusufali, & Baldonado, 1998), but for the task of identifying online sexual predators in electronic messages, computers have been shown to have accuracy greater then 94%!
"Our distance weighted k-NN classifier reaches an f-measure of 0.943 on test data distinguishing the child and the victim sides of text chats between sexual predators and volunteers posing as underage victims." (Pendar, 2007)
This compares to an accuracy of less than 85% for human experts:
"One of the coders achieved an overall accuracy of 80.40%; the accuracy for the other coder was 83.43%" (McGhee, Bayzick, Kontostathis, Edwards, McBride, & Jakubowski, 2011)
And unlike humans, computers never tire or loose focus:
"This is particularly true for long transcripts, when humans are likely to become fatigued after reading pages of chat dialog." (McGhee, Bayzick, Kontostathis, Edwards, McBride, & Jakubowski, 2011)
The figures above relate only to the use of linguistics in the classification of documents for the detection of sexual predators. Artimys goes above and beyond this by examining additional dimensions of electronic messages to detect threats. Artimys employs Sentiment Analysis and Psychometric Analysis to search the emotional and personality content of the message. This information is used as an additional “fingerprint” to uniquely distinguish threats from non-threats. It’s been shown this type of analysis alone can yield high accuracy:
“... classification based on the proposed features achieves accuracies of up to 94% ...” (Bogdanova, Rosso, & Solorio, 2012)
The combination of linguistic, personality and emotional patterns produces an unequaled level of accuracy and protection for your children.
Artimys uses several advanced technologies to protect children online including Machine Learning, Psychometric Analyis, Sentiment Analysis and Natural Language Processing. Learn more about these technologies ...
Sentiment Analysis is the measure of emotion in text. The Artimys Guardian Service uses Sentiment Analysis to measure dozens of emotions and combines these emotional measures to form an additional fingerprint (along with linguistic and psychometric fingerprints) to uniquely identify threats in text messages.
Psychometrics is the study of psychological measurement. The Artimys Guardian Service uses psychometrics by translating word choice and usage into dozens of psychological measures. These psychological measures are combined to form a fingerprint (along with linguistic and emotional fingerprints) that uniquely identify threats in text messages.
Natural Language Processing (NLP) is a broad field of Linguistics and Computer Science that focuses on machine understanding of human language. The Artimys Guardian Service uses NLP to handle slang, emoticons, spelling errors and synonyms effectively.
Machine Learning is a branch of Artificial Intelligence that deals with automatic pattern recognition by learning from examples. Artimys has a large database of message threads. Some of the message threads contain actual predators attempting to prey on children. The other message threads are teens and adults chatting about common every day events. Artimys uses Machine Learning to examine these real message threads to learn the unique patterns of the predator threads. We then use these patterns to evaluate how likely it is that your child is communicating with a predator.
Bogdanova, D., Rosso, P., & Solorio, T. (2012). On the Impact of Sentiment and Emotion Based Features in Detecting Online Sexual Predators. WASSA '12 Proceedings of the 3rd Workshop in Computational Approaches to Subjectivity and Sentiment Analysis , 110-118.
McGhee, I., Bayzick, J., Kontostathis, A., Edwards, L., McBride, A., & Jakubowski, E. (2011). Learning to Identify Internet Sexual Predation. International Journal on Electronic Commerce , 15 (3), 103-122.
Mehran, Yusufali, S., & Baldonado, M. (1998, June 23-26). SONIA: A Service for Organizing Networked Information Autonomously. Proceedings of the third ACM conference on Digital libraries , 200-209.
Pendar, N. (2007). Toward Spotting the Pedophile Telling victim from predator in text chats. International Conference on Semantic Computing, 2007. , 235-241.
Sebastiani, F. (2005). Senior Research Scientist. Text Mining and its Applications , 109-129.
Artimys Language Technologies is a technology company with deep expertise in intelligent, online child-protection products. With expertise in Linguistics, Artificial Intelligence and Data Processing, Artimys created a new category of online security for children and the parents who care about them.