The Intervention Journey A Roadmap to Effective Digital Safety Measures 2025
Page 23 of 45 · WEF_The_Intervention_Journey_A_Roadmap_to_Effective_Digital_Safety_Measures_2025.pdf
Identification
The prevalence of CSEA online is an ever-present
and growing issue. If interventions can be made
early enough, then LFF believes that users can be
stopped through intervention. Services that LFF
and other NGOs provide around the world include
brands such as Stop it Now, Safe to Talk and
Talking for Change.
The objective was identified as preventing searches
and attempted consumption of CSEA material.
Reducing demand reduces supply and therefore
prevents such abhorrent material from being
created in the first place.
Deterrence messages for those searching for CSEA
material and other harmful content were already in
place on Aylo’s adult entertainment platforms, who
worked directly with LFF and similar organizations
around the globe to provide the messaging and the
local resources to affected users. The IWF and LFF
sought to try other ways of signalling to such users
to seek help to change their behaviour.
Design
The IWF created a working group to help devise
the chatbot and ensure expert input into its
development. The group met frequently throughout
the months leading up to the release of the
intervention to ensure all voices were heard and
expertise was considered. Frequent discussions
took place within the working group throughout the
development of the chatbot. Focus groups were
used to evaluate the prompts and responses.
Partners in the design process beyond the IWF
also include LFF and The University of Tasmania
(UoT). All three organizations provided anonymous
use data to UoT, which analysed it in detail and
produced a public report on its success. LFF
created the prompts and responses for the chatbot
and tested them in focus groups.
The combined cross-sector working group proved
very successful in ascertaining how the chatbot
should function and what metrics were to be
analysed. The varied pieces of expertise provided
by each party allowed the project to develop at a
high standard and achieve its goals.
Implementation
The reThink project aims to minimize the demand
for CSEA material and to make users aware of the
Stop It Now! support service when they attempt to
search for CSAM on a legal pornography website in
the UK. The reThink intervention commenced in February
2021. A warning message was displayed on one of
Aylo’s online adult entertainment platforms in the UK
whenever a user searched for a term that Aylo had
designated as potentially relating to an interest in
CSAM. This warning message was displayed for
12 months, unchanged.
On 11 March 2022, the warning message was
supplemented with a chatbot that had been
developed by IWF. The chatbot was a simple
conversational agent built on the Google Dialogflow
ES platform. The chatbot allowed users to click on
buttons to select a path forward to information or
to manually enter text. The chatbot functionality
was designed to minimize the risk of inappropriate
responses, and provided an efficient way for users
to quickly receive information about the support
services available. Responses are predefined, and
not made using generative AI.
The chatbot conversations were monitored
throughout the length of the intervention, which
enabled it to be continuously improved to ensure
it was as effective as possible at connecting
users with support services. This included adding
additional responses and modifying how the
chatbot operated to ensure appropriate responses
were provided to users.
The original warning message operated
uninterrupted and unmodified for 14 months, after
which it was modified twice to trial different wording
and to test the effect of removing the chatbot.
The warning message was modified on 11 May
2023, and then again on 8 June. The chatbot was
disabled from 6 July through to 3 August. Data
collection ended on 31 August 2023.
Feedback, measurement
and transparency
User feedback was extremely limited due to the type
of intervention and the users it was targeting, though
responses entered into the chatbot by users are
being used for future developmental changes. The
entire project was evaluated by a third party – UoT –
and their detailed analysis is available online.13
In evaluating whether the reThink chatbot was
effective, the following metrics were analysed:
–Number of sessions containing CSAM-related
searches
–Total number of CSAM-related search queries
–Number of chatbot views per session
–Number of chatbot triggers per session
–Search types after chatbot trigger Reducing
demand reduces
supply and
therefore prevents
such abhorrent
material from being
created in the
first place.
The Intervention Journey: A Roadmap to Effective Digital Safety Measures
23
Ask AI what this page says about a topic: