TABLE OF CONTENTS

Introduction


Warning: There are examples of explicit content in this article!


We have introduced a Content Filtering feature, which allows parents to be notified when their child is sending and receiving explicit content through text and image messaging. We use an artificial intelligence (AI) system to detect when such content is being sent and received. The parent has the option to be notified of such activity, and to blur the explicit content being sent to the child.


1. Navigating to the Content Filtering Settings Page


From the Parent Portal homepage, navigate to 'Settings' and then 'Content Filtering.' 



2. Turning on Content Filtering and Adjusting the Settings


Once you are on the Content Filtering page, toggle Content Filtering ON.

From here, you can customize the settings. Please see the screenshot below, along with the corresponding descriptions of each feature!

Blur Flagged Content

If you have this ON, the flagged content would appear as three asterisks (***) on the child's device. If it is OFF, the flagged content would show up on the child's phone as normal. Either way, the flagged content would appear in the Parent Portal for the parent to view. 


Flagged Content Email Notification Frequency

You have the option to get email notifications when the child is sending and receiving flagged content. You can choose to get these notifications as soon as possible, daily, weekly, monthly, or never. 


Add or Remove Specific Words to Flag

If you click the purple 'Manage List' button, you will have the option to create a list of words that you always want to be flagged, or allowed. For example, the AI wouldn't flag the word 'house,' but you can list it as a word that you would like to be flagged. On the other hand, the AI might flag a swear word, and you can choose to allow it. 


Filter Preferences

There are four different filters that can be edited through the Parent Portal. The scale ranges from 4 (child) to 18 (adult). The lower the age, the more content is flagged. The higher the age, the less content is flagged.

  • The filters include:
    • Profanity (i.e. swear words)
    • Suggestive (i.e. sexual content)
    • Violence/Gore
    • Drugs/Alcohol
    • Suicide - There is a fifth filter that is not able to be turned on or off in the Parent Portal. If the child is sending suicidal language, the AI should pick up on this and mark it as flagged. However, if they are receiving it, it would not get flagged under the Suicide section.


You might be wondering if there is a customer-facing list of flagged words associated with each number on the scale. There is not one, but you will learn later in this article how to train the AI to tailor to your specific needs. 


3. Viewing Text Messages

On the main dashboard, take a look at the Text Usage box. You will see the number of Messages, Flagged messages, and Blocked messages. If the message is flagged, inappropriate content has been detected. If it is blocked, this means an unapproved contact attempted to text the child. If the time scheduler is enabled and a message is sent during the blocked off time period, the message would also appear in the blocked section. Once the blocked off time period is over, the message would then go through to the phone, although it would still appear as blocked.


To read the messages, click on 'See More' or click one of the contact's names listed underneath. Let's see what the child (Bill) is up to!  

                                                 


You will notice four sections.


Section 1 - General Dashboard

  • You will have the ability to change the date filter. You can choose to view content from that day, the day before, the last 2 days, the last 7 days, and the last 30 days. 
  • By clicking on 'Messages,' 'Flagged,' and 'Blocked,' you can view the total number of messages sent and received in the selected time period, as well as the number of flagged messages sent and received.
  • You are able to see the most frequently used words sent by the child (not including words such as 'the,' 'and,' 'in,' etc.) in the selected time period.
  • You can view the most frequent senders during the selected time period. You can see how many messages were sent and received, and how many of those messages were flagged. You can click on the bar next to the contact's name to be brought to text message chain with that contact.

Section 2 - Text Message Viewer

  • The text message viewer mimics the view that the child sees from their device. You are able to see a list of the text conversations from the selected time period and read through each one.
  • By clicking on the filter logo, you can filter based on sent/received status, message type, and flagged status.
  • If you are viewing a specific text message chain, you can click the orange graph icon (right underneath the arrows). You will be able to see statistics regarding that specific text chain, including most frequently used words, received messages, sent messages, and media messages.
  • You can search for keywords in the text message chain.

Section 3 - Image Viewer

  • The image viewer allows you to see images that have been sent and received during the selected time period. If you click on an image, it will bring you to the point in the conversation where the image appeared, for context. You can search a contact's name in the search bar to identify which pictures were sent to/received by that contact.

Section 4 - Message Viewer

  • You can view the time periods that texts are being sent and received. You will be able to see how many were flagged and how many went through as normal messages.

4. Understanding the AI and Flagged Content

In the previous image, you probably noticed red boxes around some of the text messages/images. These red boxes mean the content was flagged. By clicking on the text message or image, you can learn why the content was flagged, and train the AI.


You may see that some of your Content Filtering Preference bars start to shift a little bit to the left or right. For instance, the Profanity Filter Preference might be at 9, even though you originally selected for it to be at 8 before you trained the AI to filter content to your specific needs. 


            Before:

            

            After:


Please see the below examples to learn how to train the AI!


Example A

The AI flagged the below text message due to an inappropriate word and emoji (to some, an eggplant can have a different meaning!). You can tell the AI to flag texts like this less often, flag texts like this more often, or let it know that the text was flagged correctly. It is important to click 'Save and Close' after making your selection.


Example B

In the below example, the text message did not get flagged. You can either:

  • Keep the bar at the current number and train the AI by telling it to flag texts like this more often.
  • Lower the number on the 'Drugs/Alcohol Filter Preferences' bar that was mentioned previously.


Example C

You noticed there was a flagged photo, so you read the text conversation, for context. You may find that images like this are appropriate for your child to see, so you tell the AI to flag messages like this less often.




Example D

You may be wondering why the message below is flagged. Earlier in this article, we decided to always flag the word 'house.' Putting a word in the list to be always flagged (or always allowed) will override the AI. If the word is in the list of words to be flagged but the AI doesn't think it should be flagged, it will still be flagged.



5. Device View

If you choose to blur flagged content on your child's device, you will see that there are three asterisks on content that they receive, but content they send out will appear as normal. 


Below, you will see the recipient's view. Flagged words that were sent from the Troomi device will appear as three asterisks, and words that were sent to the Troomi device will appear as normal when being sent, although they will appear as three asterisks on the Troomi device.