In the UK, Ofcom has published two economic discussion papers targeted at building knowledge and understanding of safety measures, sharing best practice and developing its understanding of how it might best evaluate online safety measures under the UK's Online Safety Act 2023 (the Act).
For anyone subject to the measures under the Act, it makes interesting reading and provides insights into Ofcom's developing thinking in relation to its approach to the evaluation of effectiveness of measures adopted by services under the Act.
Evaluating online safety measures
The first economic discussion paper is a general paper titled Evaluating online safety measures - Ofcom (May 2024).
It sets out an evaluation framework that could be used to assess the effectiveness of safety measures to mitigate against the risks of online harms. It outlines a four-stage process which can be summarised as follows:
Stage 1: Objectives and appraisal
Services should understand the purpose of the safety measure and how it will achieve its objective.
Ofcom's model anticipates that there will need to be consideration of interdependencies, feedback, user experience arising from the change, external factors and unintended consequences in order to be effective.
The paper outlines an approach called "Theory of Change" which provides a roadmap for evaluating the effectiveness of safety measures over time. Any worked example using the Theory of Change model will be more complex than the below illustration, but it does provide useful checkpoints for anyone creating or reviewing an evaluation framework:
The paper refers to understanding and consideration of "unintended consequences" as being important in the evaluation cycle. One example is that a "safety measure may reduce risks of harm to a number of users, an unintended impact could be greater harm for some users (e.g. due to users that seek to bypass the safety measure and as a result end up encountering more harmful material). Therefore, the net effect on users will depend on the scale of both intended and unintended effects."
Stage 2: Identification of success metrics
Using the roadmap (likely to be more complex than the above figure – for a worked example see Figure 3 in the paper which is not reproduced here), the paper suggests that services should quantitatively and qualitatively measure progress toward these objectives using key success indicators and relevant metrics.
Indicators might include metrics like user engagement with safety features or rates of reported harmful content. Metrics likely fall into three categories: Process, Risk and Harm, with the potential for an "unintended outcomes metric" in some cases. In order to measure the metrics, the paper suggests using the S.M.A.R.T methodology (smart, measurable, achievable, relevant and timebound).
Stage 3: Assessing outcomes
In interpreting the metrics, the paper outlines four methodologies (quantitative comparisons and analysis, econometric techniques, A/B testing and behavioural techniques).
Stage 4: Evaluate and iterate
Regular analysis of collected data ensures services will be able to assess the effectiveness of their safety measures.
As anticipated by my colleagues Lorna Cropper and Frankie Everitt in their blog How Ofcom's approach to Video Sharing Platforms will inform its regulation of the Online Safety Act, Ofcom is using real world examples of video sharing platforms to develop an approach to effectiveness under the Act. Part 2 of the paper contains a helpful worked example of the model in the context of Video Sharing Platforms and parental controls.
Don't miss a thing, subscribe today!
Stay up to date by subscribing to the latest Public and Regulatory insights from the experts at Fieldfisher.
An evaluation of the impact of Twitch's content classification labelling
The second paper is a case study titled An evaluation of the impact of Twitch's content classification labelling - Ofcom (September 2024). The paper examines changes made by Twitch to its content classification labelling (CCL) in 2023.
Twitch introduced changes to its content classification guidelines which moved from a binary mature/non-mature classification to six CCLs (mature-rated game; sexual themes; drugs, intoxication or excessive tobacco use; violent and graphic depictions; significant profanity or vulgarity; and gambling). The changes included penalties for content creators that failed to accurately label streams, including labelling the creator's channel with the relevant CCL for a set period of time irrespective of whether all streams fell within the definition for the CCL.
The findings can be summarised as follows:
- content labelling accuracy by content creators increased substantially following the CCL change;
- the CCL change did not materially alter the type of content produced by creators; and
- the CCL change did not alter viewer behaviour materially.
The paper underscores the complexity of online safety and the need for comprehensive evaluation frameworks that incorporate various metrics and stakeholder inputs. It provides a step-by-step approach that could be used to evaluate online safety measures and signals that Ofcom is more likely to expect in-depth analysis rather than "simplistic comparisons of high-level metrics". Finally, the case study highlights that the changes did not substantially change what creators are developing or what this set of viewers were watching.
What should services do with this information?
- Familiarise yourself with the Act and understand where your service may be caught by the definitions within the Act.
- Once the Codes of Practice are in force, conduct the risk assessments under the Act relevant to your service. There is likely to be some work which could be carried out in advance of this date and services should start the review exercise as early as possible. The illegal harms code of practice remains in draft (it is expected to be published by the end of the year) and the draft of the children's access assessments code of practice is awaited. In the interim, the discussion papers provide examples of good practice that services may find useful when assessing compliance with the Act and future codes of practice. The ICO's "Age Appropriate Design: a code of practice for online services" has been in force for some time and should also be considered.
- Understand the potential harms associated with your platform through risk assessments.
- Develop and implement measures designed to address these risks. This could include content moderation, reporting mechanisms and user support systems and tools. User and other stakeholder feedback can also be helpful in understanding the effectiveness of these measures and facilitate continuous learning and improvement.
- Data is key: robust data collection is likely to be key to evaluating the safety of online safety measures.
- Review and refresh: set regular reviews of the systems (at least once a year and any time you make a change to the service) which consider the efficacy of the system and improvements.
- Implement a clear system of governance which ensures regular reviews and reporting at a senior level within your organisation.
- Be prepared to respond to requests from the regulator which seek to understand the measures that you have in place for evaluating online safety and how you carry out such assessments and the continual feedback loop for learning and improvement.
Together, the papers are over 100 pages long and targeted at generating discussion from services, academics, civil society organisations and the broader trust and safety and online safety expert community.
With the staged approach to implementation of the Act, evaluation of safety measures may not be top of most services' agendas but, for those interested in looking forward to what regulation might look like and seeking an opportunity to engage with its development, the papers are informative reading.
For advice on how the Act may impact your business, or for guidance on compliance and evaluation, please contact John Brunningand Nicola Margiotta.
Our thanks to Lorna Cropper for her assistance in the preparation of this blog.
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.
