Towards sustainable short-form video: Modelling solutions for social and environmental challenges

A. Social impacts of short-form video

From a social perspective, short-form video applications can have both positive and negative impacts. One major argument in favour of short-form video is its use as a learning tool, involving a range of education-oriented stakeholders, such as charities, universities, and professionals.^2,30 Educational initiatives on short-form video platforms have been applied to a variety of topics, including science,³⁰ physical health,³¹ mental health,³² and climate change,^33,34 achieving moderate success. However, these initiatives often face the issue that most shared information does not originate from reputable sources.^31,33–35 Even in the case of mental health professionals, condensed social media posts on mental health topics may include inaccurate information.³⁶

Early research suggests certain mental health benefits of these applications, such as an increased sense of community, validation of experiences, and facilitation of dialogue. Nevertheless, there are significant limitations and drawbacks in this regard.³⁵ Other studies indicate that positive mental health outcomes are achievable if users actively seek out appropriate content types.^37,38 The extent of these benefits, however, remains uncertain.³²

While there is insufficient evidence to fully assess the overall effect of short-form video on social aspects like mental health, a growing body of literature highlights the risks associated with its use.

Several researchers argue that the design of short-form video applications incorporates notably addictive features,^12,13,39 leading to measurable addiction rates, particularly among younger populations,^10–12 and significantly impacting mental health outcomes.^12,40 Social media addiction is estimated to affect over 17% of the global population,¹² with different estimates provided for problematic use on platforms like TikTok. Depending on the user group and dataset, estimates range from 36% of parents reporting that their children are addicted to TikTok, to 70% of young users who consider themselves addicted to TikTok [11]. Other sources report different, yet equally concerning statistics [12]. Due to the severity of this issue, the TikTok Addiction Scale¹² was developed to study addiction to TikTok specifically.

Use of short-form video platforms, especially when excessive, has been directly linked to higher rates of depression, anxiety, loneliness, and other mental health problems across various populations, with findings consistent across age groups, genders, and geographic regions.^5–9 Some suggest that these effects are intensifying over time.⁹ These effects are especially pronounced among younger users, who tend to consume this type of media the most.^7,35 A contributing factor is the prevalence of self-harm and suicide-related content on platforms like TikTok, as well as the spread of misinformation on mental health topics.⁶ Despite the real mental health challenges many short-form video users experience, finding a solution may be hindered by self-diagnosis behaviours, leading to inaccurate self-assessments based on unreliable information from non-reputable sources.⁴¹ This increases the risk of misdiagnosis and avoidance of professional help.⁶

Body image issues are a significant consequence of the current usage patterns of social media.⁴² Since major social media providers use short-form video as a primary content format, this problem has been directly linked to applications such as TikTok and associated with mental health issues such as eating disorders.⁶ Studies indicate body image impacts on both male⁴³ and female³⁸ users.

Another risk factor is the endorsement of potentially life-threatening ‘challenges’ through viral trends on these platforms. A notable example is the death of an adolescent girl who participated in the “Benadryl Challenge”, ingesting a fatal dose of the allergy medication Benadryl.¹⁶ Another life-threatening trend was the “Labello Challenge”, where users would remove pieces of a lip balm stick when feeling sad and engage in self-harm when the stick was empty.⁴⁴ These trends are rising in popularity and represent a serious public health concern.⁶

The spread of misinformation and harmful content through short-form video applications is also on the rise, posing further societal risks.^14,15 As the use of these applications increases, so does the likelihood of vulnerable individuals encountering misleading or dangerous content. Addressing misinformation on short-form video platforms is critical to societal well-being; however, this is a challenging task, and debunking campaigns have been effective only to a “modest” degree, often resulting in the opposite effect.¹⁴ The risk of spreading dangerous information cannot be ignored either. As described in,¹⁵ “TikTok users are sharing calls for violence against people of color and Jews, as well as creating and sharing neo-Nazi propaganda.” The situation is especially serious as their analysis reveals that TikTok’s efforts to eliminate content containing threats, violence, discrimination, and even terrorism promotion, are largely ineffective. Given the high number of vulnerable underage users on these platforms,⁶ failing to address the presence of “paedophiles, terrorists and extremists” could have devastating social consequences.¹⁵

Despite uncertainties about the overall impact of short-form video applications on their users, the substantial evidence of the risks associated with short-form video consumption highlights the need to develop and implement solutions to mitigate these risks. In line with calls from the literature for regulation and interventions,^6,10,13,37 one of the objectives of this research is to provide and analyse viable solutions to reduce the negative social impacts of short-form video.

B. Environmental impacts of short-form video

In the discussion about the impacts of short-form video, the environmental aspect is often overlooked. This is evidenced by the scarcity of studies identified during the literature search conducted for this research. Addressing this gap is vital for three reasons.

First, there is a general lack of awareness about the environmental impacts of video streaming.¹⁸ Second, due to its high popularity, short-form video is a major contributor to internet traffic, with estimates suggesting it drives the total share of IP traffic attributable to video content to over 90%.¹⁹ Finally, the environmental impacts of this rapid growth are further exacerbated by the unique characteristics of its format, such as the fact that not all downloaded and cached content is consumed,⁴⁵ leading to considerably higher emissions than regular streaming.¹⁸ For example, behaviours like fast video switching in short-form video consumption result in approximately 45% of mobile data losses.²⁰

Researchers typically approach short-form video as part of broader streaming or social media studies. According to a report by Cisco [13], video streaming is a major contributor to overall internet emissions. Several studies corroborate this claim, emphasizing the substantial environmental impact of video streaming and the critical importance of mitigating it.^18,19,46,47 However, these analyses often fail to address the specific impacts of short-form video streaming from a systemic perspective, potentially resulting in an incomplete understanding of its environmental effects.⁴⁸

While streaming can be a more sustainable alternative in certain contexts, it poses several potential challenges.¹⁹ To address these challenges and mitigate streaming’s environmental consequences, some authors adopt behavioural approaches. For instance,⁴⁹ examines the willingness of users to compromise video quality to conserve energy and reduce costs. The authors of¹⁸ investigate the effects of raising awareness about streaming’s environmental impacts to promote more sustainable user practices. In,⁵⁰ the authors analyse the effects of modifying luminance and resolution on both user experience and sustainability outcomes.

A study specifically applied to short-form video is presented in,⁵¹ which evaluates user responses to various streaming experiences and proposes adapting transmission based on metrics of acceptability and annoyance. Another relevant example is,⁵² where the author introduces design features aimed at reducing the environmental impact of the Instagram app. While the study does not focus exclusively on video streaming, a significant portion of Instagram content comprises short-form videos, and the proposed measures directly affect the environmental performance of video streaming.

On the other hand, some authors adopt a more technical approach, focusing on optimizing the technology used for streaming video content. The authors of⁴⁵ propose an alternative content transfer model designed to adapt to the bursty nature of short-form video while meeting its low-latency and high-bandwidth requirements. In,⁵³ significant energy improvements are achieved by optimizing short-form video compression, minimizing video quality loss in the process. The authors of²⁰ introduce a system that leverages previous network data to develop intelligent adaptation models, achieving savings of over 80% in mobile data usage. While these technical solutions are crucial in addressing the environmental impacts of video streaming, the present study takes a different focus and does not delve into the intricacies of implementing such approaches.

Studies exploring the connection between environmental sustainability and social media often emphasize the use of social media as a tool for promoting sustainability. Some authors report promising results, showing that short-form video content creators achieve greater communication efficacy compared to other media, highlighting potential for more impactful climate campaigns.^54–56 In a related study,³³ the authors review 100 popular TikTok videos on the topic of climate change, collectively amassing over 205 million views. Their findings suggest that, while platforms like TikTok can reach vast audiences, the absence of reputable sources undermines the accuracy of the information shared in these videos. Despite being a relatively recent study, the analysis conducted in³³ is among the first of its kind, underscoring the need for more research on the interplay between social media and environmental sustainability. Addressing the lack of reputable sources is particularly challenging, as public opinions on topics like climate change are often more influenced by social media influencers than by scientists or institutions.³⁴ This is not unexpected, given the demonstrated effectiveness of influencer marketing strategies in driving consumer behaviour.^1,57

To deepen our understanding of the environmental consequences associated with the rise of short-form video applications, it is essential to extend beyond the research mentioned above.

In a series of studies conducted by France’s National Agency for the Ecological Transition and Arcep,^58,59 the current sources of the environmental footprint of digital technologies and their future consequences are analysed. Their findings indicate that user devices are the primary contributors to the environmental impacts of technology, accounting for between 65% and 90% of these impacts. This footprint primarily arises from energy consumption during both production and utilization, as well as the depletion of natural resources. More specifically, their results identify screen use as the largest single source of impact. The studies also emphasize the significant role of video streaming in exacerbating the environmental footprint of digital technologies. Based on their analysis, they advocate for prioritizing measures aimed at reducing screen time and extending device lifespans as central to discussions on sustainable technology.

Device use and screen time, however, are not static measures; current trends indicate that people are spending an increasing amount of time on their devices, particularly smartphones. Device usage time has risen steadily [14], especially among younger generations, who are the heaviest users of devices, primarily for accessing social media. Social media applications, particularly short-form video platforms such as TikTok, are at least partially responsible for this increase in device usage [15]. This trend is logical, given the rapid growth of short-form video content and its engaging, often addictive nature.^10,12,13 Data shows that the average daily time users spend on TikTok has doubled between 2019 and 2023.¹⁶ Since the growth in short-form video consumption contributes to an increased device use, also increasing screen time, it directly results in an increase in the primary source of impact of digital technologies.

The forecasts from the ADEME–Arcep study⁵⁹ emphasize that extending device lifespans is crucial for achieving sustainable information and communication technologies [16]. Interestingly, none of the study’s scenarios predict a decrease in device lifespan; their baseline scenario assumes it will remain constant in the future. However, it is possible that device lifespans may shorten, in part due to increased short-form video consumption.

The rise in short-form video usage not only increases energy consumption—especially in comparison to other streaming activities—but also results in more frequent charging of devices. Frequent charging accelerates battery degradation, reducing their useful life.⁶¹ Since battery performance is a key factor influencing customers’ decisions to replace their smartphones,⁶² this dynamic could lead to more frequent smartphone purchases. Additionally, the shift from consumers to prosumers⁶³—where users actively create as well as consume content—could reshape the features that individuals prioritize in smartphones. For example, over 50% of TikTok users actively engage in content creation,⁶⁴ potentially driving demand for devices with high-quality cameras, a feature already highly valued in smartphone purchases.⁶⁵ Another factor influencing smartphone demand is platform-induced ecosystem switching. Some users may opt to switch from one smartphone ecosystem to another to optimize their experience with specific social media applications like TikTok [17], even if their current device is still functional. Furthermore, platforms such as TikTok promote consumerist behaviours, further amplifying these trends.⁶⁶

To fully assess the environmental impacts of short-form video, it is also essential to consider data centre usage. According to the ADEME–Arcep study,⁵⁸ data centre usage constitutes the second-largest source of digital technologies’ environmental footprint, following user devices, with network usage ranking third. The rapid growth of short-form video increases these impacts due to its high bandwidth requirements,⁴⁵ direct data losses and inefficiencies,²⁰ and reliance on processing personal data and algorithmic recommendations.⁶⁷ These factors collectively intensify the environmental impact of shortform video, particularly when compared to other content forms, including long-form streaming.¹⁸

Some of the mechanisms outlined in this section to highlight the environmental consequences of short-form video consumption remain theoretical, as no direct evidence currently establishes causality. This is particularly true for the links between increased short-form video consumption, increased device usage, and reduced device lifespans. However, the mechanisms described represent plausible connections and offer a foundation for future research. This is especially important given that no prior studies have analysed the environmental impacts of short-form video from a perspective similar to the one presented here. Coupled with the scarcity of systematically collected data on these links, this highlights a significant research gap that requires urgent attention.

It is also important to note that even if future research fails to establish direct causal relationships, the core environmental impacts discussed in this research would remain largely unaffected. For instance, while short-form video may not directly increase device usage, existing data confirm that average device usage time is rising. Concurrently, the time people spend consuming short-form video content is growing rapidly, particularly among younger generations. Given that short-form video has a disproportionately large environmental impact compared to other forms of content consumption, it is still a significant contributor to the environmental impact associated with personal device use and data centre operations. This remains true even if future studies reveal that short-form video is not the primary driver of increased device usage.

Similarly, if further research finds no substantial link between short-form video consumption and reduced device lifespans or increased smartphone purchases, the data presented still suggest that current consumption trends hinder the necessary improvements in device lifespans. Extending device lifespans is essential to achieving sustainability goals in the information and communication technologies sector.⁵⁹ Addressing the challenges posed by short-form video consumption is crucial, as these could slow progress toward this goal. This conclusion would remain valid even if further research indicates that short-form video does not significantly contribute to device lifespan reduction or encourage increased smartphone purchases.

C. Compartmental Models for Short-form Video

The population dynamics simulations used in this research are performed using a compartmental model. Compartmental models, first introduced in epidemiological research in 1916,^68,69 are a powerful tool for analysing the dynamics of evolving populations. These models divide a population into distinct groups, or compartments, and represent the movement of individuals between these compartments through differential equations. Over time, this modelling approach has been successfully adapted and expanded to diverse fields, including market dynamics in the digital economy.^63,70

There are several reasons for choosing this modelling strategy in this study. First, as stated in Section II-B, research on shortform video lies at the intersection of video streaming and social media research. Recent studies suggest that compartmental models are effective in both contexts. These models have been employed to simulate the effects of environmental policy implementation and competition on video streaming platforms,^47,71,72 as well as in the context of music streaming.⁷³ They have also been applied to studying the evolution of social media applications,^74–76 offering an effective framework for simulating phenomena such as network effects while providing a more streamlined implementation compared to alternative techniques.⁷⁵ These applications suggest that compartmental models are well-suited for the study of short-form video applications.

Network effects, defined as the increase in the perceived value of a product or service due to the growing number of its users,⁷⁷ are a critical factor in understanding the growth of social media platforms.⁷⁸ An early example of network effects modelling is the Bass model,⁷⁹ which describes how the purchase of durable goods is influenced by the number of prior sales, making customers more likely to buy a product as more units are sold. Several models have since been developed based on the Bass diffusion equation, describing the evolution of complex economic systems, such as the adoption of online video games or forecasting social media usage. Notable examples include the Buyer-Player-Quitter model⁷⁴ and the PHLoQui model.⁷⁵

All models directly based on the Bass model operate under the implicit assumption that Metcalfe’s law accurately describes the evolution of a network’s value.⁸⁰ However, this assumption has been contested, with researchers suggesting that social media networks are better characterized by Odlyzko-Tilly’s law.^63,81 Subsequent studies have revisited this debate, reaffirming the applicability of Metcalfe’s law in various contexts,⁸² including social media platforms.⁸³ Some researchers have argued that Metcalfe’s law performs better than alternative models for social media adoption,⁸⁴ while later investigations have presented contradictory findings.⁷⁶ Given the absence of prior research examining the impacts and evolution of shortform video through the methodology applied in this study, the simulations presented in Section VI incorporate both Metcalfe’s and Odlyzko-Tilly’s laws.

To mitigate the negative consequences of short-form video consumption, as discussed in sections II-A and II-B, it is essential to address problematic patterns of use. For this purpose, in addition to the digital economics literature, it is valuable to explore the application of compartmental models within the addiction literature. A systematic review of addiction models⁸⁵ concluded that “the spread of social contagious issues can be modeled accurately with epidemiological methods.” The authors argue that, despite certain limitations, compartmental models provide an effective framework for modelling population dynamics. They further note that alternative approaches, such as agent-based models, are “not yet mature regarding addiction”.⁸⁵

The population model employed in this study was developed through a careful review of existing compartmental models in the contexts of both social media usage and addiction. The adapted model incorporates the categories of early adopters, moderate and heavy users, and quitters, drawing on insights from addiction models.⁸⁶ Simultaneously, it integrates elements from customer dynamics models, such as the Buyer-Player-Quitter model⁷⁴ and the PHLoQui model.⁷⁵ This hybrid approach offers a more nuanced understanding of the dynamics at play compared to relying solely on aggregate values or total user numbers, as done in other studies.^63,73

A. Money: Financial restrictions

This scenario primarily comprises various types of taxes and restrictions based on different metrics, used as tools to moderate the impacts of short-form video. These types of restrictions require intervention from authorities and governing bodies and may encounter pushback in contexts where policymakers tend to oppose increased taxation.

This scenario includes:

B. Content: Information campaigns and content creation

A less invasive strategy for reducing the impacts of short-form video applications involves content creation and campaigning. These strategies have shown moderate success in influencing user behaviour,^18,54–56 promoting healthier and more sustainable consumption patterns, and increasing the availability of high-quality educational content.¹⁰ However, these methods face the challenge of balancing outreach with information accuracy. Social media influencers can reach wider audiences and have a greater capacity to motivate behavioural change compared to institutions and scientists.³⁴ This raises numerous challenges in addressing the lack of reputable sources on social media.^24,33,36

Effective campaigns require the collaboration between institutions, scientists, and social media personalities.³⁴ If proper precautions (see analysis in²⁴) are implemented, these content campaigns offer a significant advantage, namely, network effects. Network effects are a powerful driver in spreading information on social media and increasing platform usage.⁷⁸

This scenario includes:

C. Design: User Experience and Service Design Guidelines

This group of measures focuses on user experience and user interface design, addressing topics such as algorithmic recommendations and behavioural nudges. These approaches represent an effective strategy for encouraging both socially and environmentally sustainable consumption patterns.^13,49,52

The proposed measures can be implemented through regulations or mandatory guidelines for short-form video applications. However, one notable advantage of these measures is the possibility of establishing voluntary agreements between businesses and authorities beforehand. For example, adopting voluntary guidelines could mitigate the need for stricter future regulations, enabling platforms like TikTok to adjust their business models without additional policy intervention.

This scenario includes:

D. Combined Approach: All Measures, to a Moderate Degree

This scenario involves implementing a combination of measures from the previous scenarios. For this study, all measures outlined in sections III-A to III-C were applied simultaneously. This combined approach offers the advantage of flexibility, as measures do not need to be implemented to the same extent, enabling lighter regulations and greater feasibility while maintaining effectiveness. Although regulators may adopt a different combination of measures in practice, the purpose of analysing the set defined here is to demonstrate the benefits of a multi-pronged approach. This is particularly relevant in contexts where stringent financial restrictions are challenging to implement.

A. Conceptual model of a short-form video application

As a basis for the mathematical simulation model that we developed for this work is a conceptual model of short-form video applications that we present in this section. Arguably, the most prominent example of a short-form video application is TikTok. It is one of the largest social media platforms in existence and is entirely based on short-form video. For these reasons, data availability on short-form video activity and its effects is greater for TikTok than for other platforms. Thus, we chose TikTok as a representative case and developed our model based on its operations and business model (see Ref. 93 for a detailed description of TikTok’s business model). Nevertheless, neither the model used here nor the conclusions of this research are limited to TikTok.

Consider another platform, such as Instagram. Instagram features Reels, which function in a manner similar to TikTok. While the evolution and impacts of user interactions with Reels may not fully represent Instagram as a whole, they do reflect a significant part of its platform. Reels constitute a crucial component of Instagram’s business model, and this specific segment operates in a way that is comparable to TikTok. Therefore, although Instagram’s overall structure differs from that of TikTok, a substantial portion of its functionality can still be effectively modelled using the framework developed for TikTok.

TikTok is an example of multisided platform, defined as a digital platform offering distinct value propositions to different customer segments.⁷⁰ For TikTok, the two main customer segments are users and advertisers. TikTok provides its users with free access to personalized short-form video content and opportunities for community interaction, including both video creation and non-video interactions such as likes, comments, and direct messages. Since TikTok users often act as both consumers and producers of content, they are sometimes referred to as prosumers.^94,95 Within this user segment, a subset comprises content creators or influencers [18], who produce content professionally with the goal of earning a profit. These creators generate revenue through content monetization based on the number of views their videos receive, as well as through paid partnerships with brands.

The majority of TikTok’s revenue comes from advertisers—individuals or companies purchasing exposure through in-app advertising. These ads are seamlessly integrated into the viewing experience, often resembling regular content videos. Although the proportion of advertising within a typical TikTok session varies, some estimates suggest that between one-fourth and one-third of the content consumed is advertising [19]. Consequently, strategies designed to prolong user time on the app benefit both TikTok and its advertisers. However, the prioritization of time on app and maximum user engagement can contribute to negative social and environmental outcomes, as outlined in Sections II-A and II-B.

The focus of this study is on minimizing the negative externalities associated with short-form video consumption, particularly when usage becomes excessive. For this reason, our conceptual model centres on the evolution of user numbers and time spent on the app. The conceptual diagram illustrating this model and its feedback relationships, developed for this research, is presented in Figure 1. However, this diagram does not fully capture the complexity of the factors influencing user base evolution and time spent on the app. A more detailed analysis of these is provided in Section IV-C.

Figure 1. Conceptual dynamic model of TikTok as a case of a short-form video application.

The model shows the main actors (drawn as squares) and other main concepts (drawn as circles), along with the influences/feedbacks exercised on one another (drawn as arrows).

In the model, users can join or leave the platform at any time. Joining is influenced by TikTok’s promotional efforts, the quantity and variety of available content, and the potential for interaction with other users. User-generated data enhance personalization by improving the recommendation algorithm. The longer a user spends on the app, the more data they generate, which leads to more precise personalization. A significant portion of users also create content, further increasing the platform’s appeal. In addition to user-generated content, influencers contribute to the variety and attractiveness of TikTok’s offerings. Both content and personalization enhance engagement, driving increased time on the app.

TikTok’s business model relies on user numbers and time on app to generate revenue. A portion of this revenue is distributed to content creators (influencers), while the remainder is allocated to cover TikTok’s operating costs and investments aimed at attracting more users and improving the platform, further enhancing personalization.

B. Population dynamics model

The simulation model used in this research has been developed by adapting existing models from scientific studies on the evolution of digital platforms, social media, and addiction (as detailed in Section II-C). The model is designed to capture the key dynamics of a short-form video platform such as TikTok, and incorporates the elements and feedback loops described in Section IV-A. This user dynamics model (also referred to as the ‘users’ model), illustrated in Figure 2, consists of five compartments:

Figure 2. Diagram of the compartmental model for simulating the evolution of TikTok’s user base.

The arrows represent transitions between compartments. The parameters that affect each of the transitions are detailed in table I.

The usage time proportions defined for the three user compartments (Explorers, Light Users, and Heavy Users) represent qualitative differences and are a necessary definition from a mathematical standpoint. Alternative models might adopt different definitions using other proportions based on the qualities distinguishing the user compartments. However, this does not significantly impact the conclusions of this research, as their validity remains largely unchanged as long as the chosen model adheres to the premises defined here. In other words, the exact numerical values of the proportional usage time differences between user segments are of little importance, as long as the key quality they reflect remains consistent: habitual (light) users spend, on average, more time on the app than new, inexperienced users, and heavy users spend more time than light users.

C. Parameters of the model

To model the different factors influencing the transitions in the population dynamics model described in Section IV-B, a set of parameters is defined. Some parameters can be influenced by measures and policies, while others describe average intrinsic characteristics of users that cannot be influenced. Parameters that promote joining TikTok and/or increasing usage include:

Parameters that incentivize leaving TikTok and/or decreasing usage include:

These parameters are dynamic, as their values may change based on the evolution of the user base. They also influence transitions between compartments in different manners, as users’ behaviour and motivations are not uniform across segments. For instance, mental health concerns (m_h) may play a more significant role for heavy users leaving the platform than for explorers, who have just started interacting with it. A complete list of the parameters and the flows they affect is presented in Table 1.

Across the different scenarios from Section III, the parameters of the model reflect the influence that various measures and business decisions can have on the decisions of current and potential TikTok users, ultimately determining the number of users and their usage patterns. However, parameters such as personal reasons for joining or leaving are inherently unknown and cannot be directly modified by implementing any of the proposed measures. Moreover, as remarked in Section IV-B, the number of potential customers is also unknown, as it is not reasonable to assume that it is equivalent to the entire global population. For these reasons, a data-driven estimation is performed to assign values to these parameters, following the approach described in Section IV-E. The initial values of the remaining parameters were obtained using a sensitivity analysis based on an OAT (One-at-a-time) approach. This enabled us to determine a suitable range of values for each parameter, identifying higher and lower parameter value limits and an order of magnitude compatible with the timescale of this study.

Financial restrictions, referred to as Money measures and described in Section III-A, constitute the first scenario. These measures directly influence the amount paid in taxes and the cost of advertising. Additionally, the proposed data tax in the Money scenario affects the rate of improvement of the recommendation algorithm, thereby slowing down the development of the customization of experience parameter. Furthermore, the content-creation financial restrictions detailed in Section III-A may reduce content creation activities by prominent influencers, as measured by the novel content parameter. These seemingly minor changes significantly impact the model’s dynamics, as they affect several key elements driving TikTok’s growth within its current business model.

Information campaigns and content creation initiatives, labelled as Content measures and detailed in Section III-B, form the second scenario. Institutional and professional mental health awareness campaigns influence both the mental health concerns and external critique parameters. Environmentally-focused campaigns primarily increase the external critique parameter. While the impact of environmental campaigns may vary based on individual environmental concerns, this model assumes that the average level of’eco-consciousness’ within the population remains constant across scenarios. Incorporating the influence of a population’s environmental concerns on policy implementation, as explored in,¹⁰⁰ falls outside the scope of this study. Lastly, partnering with influencers to promote moderate consumption and leveraging network effects modifies the influencer campaigns parameter.

User experience and service design guidelines, categorized as Design measures and described in Section III-C, define the third scenario. Measures aimed at’breaking the addictive design’ of the platform directly reduce the habit-formation parameter. The customization of experience parameter is also impacted, primarily through measures targeting algorithmic recommendations. Time cues, reminders, and time limits influence the level of interference, particularly when moderate use risks escalating to heavy use. Restrictions on certain features and guidelines that may lower the quality of experience affect parameters related to new features and overall user experience.

The combined approach, corresponding to the fourth scenario, affects all parameters modified in the previous scenarios. However, these impacts are less pronounced for each individual parameter, as the measures are implemented to a more moderate extent compared to the individual scenarios. The initial values of all parameters for each scenario are detailed in Table 2. However, a full interpretation of these values requires the definitions provided in Section IV-D.

D. Mathematical implementation of the model

This section details the mathematical definitions and implementation leading to the set of equations that form the core of the simulation model. This includes all the definitions of dynamic parameters and other derived variables, as well as the representations of the impacts of short-form video consumption in the model.

Dynamic parameters respond to specific modelling needs. To describe the complex dynamics of TikTok’s evolution and impacts, static parameters are not sufficient. As presented in Section IV-A, magnitudes such as the time spent using the platform or the level of personalization of suggestions have multiple dependencies and can be influenced in different ways.

Moreover, the presence of network effects and feedback loops in the model makes this situation even clearer.

In this model, if a parameter depends on network effects, it can be defined as a constant multiplied by a network effects function F_NE. For instance, the community interactions parameter com_i can be defined as com_i = com _i0F_NE, where com _i0 is a constant that determines the relative impact of community interactions for a certain user base size. As the user base increases, the network effects function makes this impact larger. The same principle applies, in general, to all parameters affected by network effects, as presented in Table 1, with the exception of the customization of experience parameter. To model customization of experience, it is necessary to factor in the influence of revenue in addition to network effects, resulting in the definition cox = cox₀(F_NE + R), where R is the revenue at a given time (see eq. (1)). As the number of new implemented features (n_f) and advertising investment (ads) by TikTok is dependent on revenue, the value of these parameters in the model is also proportional to the revenue, while not being directly affected by network effects.

The modelling strategy for the network effects function will depend on the choice of assumptions about their nature, as described in Section II-C. Using Metcalfe’s law, F_NE should be proportional to the total number of users, while Odlyzko-Tilly’s law proposes a logarithmic relation. Moreover, the contribution of the different user groups defined in Section IV-B should not be the same, as the level of interaction, data generation, and content production varies depending on usage. Based on the rationale from Section IV-B, the ratio of contribution for the defined user segments is set to E: U = 1 : 3 and U: A = 1: 2, where E refers to Explorers, U to Light Users, and A to Heavy Users. Based on these principles, the network effects function using Metcalfe’s law is of the form

Following a similar logic, since heavier usage generates more revenue in TikTok’s current business model, Heavy Users have a larger contribution to the company’s revenue. However, a comprehensive metric of the revenue requires considering the average time a user spends on the app, and not only how much time a user spends relative to other users. For this reason, the Time on app parameter (toa), i.e., the average time a user spends on the platform on a daily basis, is defined. Since this model includes different user groups, the actual average time on app is the weighted average of the time spent by the different user types, using the proportions established in Section IV-B.

These principles allow us to define the total gross revenue as R_T ∝ (E + 3U + 6A) toa, and the net revenue as

Maximizing revenue in this model can be achieved by increasing the number of users and the average time a user spends on the platform, aiming for a larger proportion of heavy users. Analysing other sources of revenue, such as from external investors, falls outside the scope of this study.

The time on app parameter (toa) is a central element of the model. It is influenced by content novelty and diversity, as these increase the likelihood that any given user will be able to consume content that they find interesting, entertaining, or engaging in general. It is also influenced by the level of customization of experience, as an effective recommendation algorithm is key to delivering the most engaging content to each specific user depending on their personal preferences. The time a user spends on the application is also influenced by the habit-forming features of the platform and the algorithm. To account for the fact that there is a limit on the value of this variable, as well as possible diminishing returns due to increased personalization,¹⁰¹ we define the time on app parameter as

To complete the simulation model, the values of unknown parameters that cannot be modified are estimated from data. To make the model representative of the real evolution of TikTok, constants such as toa₀, the personal preferences parameters pp_j and pp_l, and the potential size of the market are obtained as described in Section IV-E.

Based on the definitions in this section, the system of equations used for simulating the evolution of TikTok, eqs. (3) to (6), is defined:

The system of equations was solved numerically in python using the explicit Runge-Kutta method of order 5(4), using the solver provided in the scipy library. The code for the simulation model is available in the online repository Zenodo [21].¹⁰²

Based on the magnitudes resulting from the simulation, the evolution of the impacts of TikTok is calculated. The definition of the environmental impact is based on the assumption that it increases with the number of users, depending on the usage intensity. Therefore, to maintain consistency across definitions, the environmental impact is measured as E_i = E₀(E + 3U + 6A) toa, where E₀ is a constant. This definition is motivated by the reasons explored in Section II-B, as well as by the approaches utilized in previous studies.^71,73,100 This assumption can also be justified in part as a consequence of the second law of thermodynamics, since data transfer always requires energy. Thus, increasing data transfer would increase energy consumption, and therefore increase environmental impact. Although the current state of global energy production is far from being 100% renewable-based,¹⁰³ this reasoning would still hold true, as renewable energy sources also have negative environmental impacts, albeit much smaller than fossil fuels.^103,104

Distinguishing between different types of environmental impact in this type of study may be useful in certain contexts.¹⁰⁰ However, in this case, we opted for using an aggregate measure to allow for an interpretation independent of the metric utilized. In practice, the comprehensive metric used in this model can be interpreted individually, for instance, as CO₂ emissions, energy use, or e-waste generated, and also as a measure of impact in general, aggregating multiple factors simultaneously. In this study, the validity of this metric is not compromised by this choice because its absolute value holds little relevance, as the results of this study stem from relative comparisons of changes across scenarios. Measuring magnitudes such as the exact amount of CO₂ emissions due to TikTok usage, or evaluating the absolute global environmental impact of short-form video in general, falls outside the scope of this study.

Defining a metric for the negative social impacts of TikTok is a challenging task, given the data and quantitative research scarcity on the impacts of digital technologies on metrics such as safety and crime, working conditions, health benefits and risks, or inequality.⁹² As presented in Section II-A, mental health issues are often highlighted as the most prominent risks that short-form video consumers face. Due to the growing body of scientific literature and recent policy developments focusing on this aspect, this model implements an estimation of mental health impacts as one impact metric.

Mental health issues arise more often due to problematic use patterns. Thus, one way of estimating the general mental health impact of TikTok is defining a metric based on the number of heavy users and the average daily usage time. In this model, the metric for mental health impact is defined as M_i = M₀ × A × toa, where M₀ is a constant. Following the same logic as with the environmental impact metric, the absolute value of the mental health impact metric is not relevant, as the results are based on relative comparisons.

Different and more advanced metrics could be developed based on the current model’s outputs, as well as future extensions and adaptations. The open availability of our model is intended to facilitate this task.

E. Development of baseline forecasts

This section presents the methodology used to define a realistic baseline for modeling the evolution of TikTok and obtain the values of various unknown parameters. The strategy consists of two stages. The first stage involves obtaining a robust data-based forecast of the size of the user base of TikTok by employing and comparing existing models that have been used in research for similar purposes. The second stage uses user base size and average usage time data to obtain the unknown parameters in this novel model, defining the baseline. The forecasts produced by the new model are compared to those from the first stage to verify that the projected final size of TikTok’s user base in this new model is compatible with those more robust estimates. The results of this process are presented in Section V.

The data employed in both stages of this process [22] (quarterly total active users data from the year 2018 to the year 2024) was selected because it is openly available and offers higher granularity than other open datasets. While other sources [23] were considered too, they were not used because of access limitations and lower granularity. Even though different sources provide somewhat different estimates of the size of the user base at any given point in time, all of the sources considered portrayed a similar and consistent evolution trend, and the differences we encountered did not exceed 10% with respect to the data chosen for this study. Lastly, the data used in this study offered the most conservative estimate of the number of TikTok users, measuring only active users rather than registered accounts. Using these data contributes to reducing the risk of overestimating the potential size of the platform and the impacts of its evolution.

Given that all of the models used in this study are freely available, it is possible to directly apply them to obtain other forecasts based on different datasets. However, the conclusions of this study would not be altered in any significant way if a different dataset was chosen for the definition of the baseline. This is due to the fact that the general evolution trend remains consistent across datasets and the conclusions are mainly based on relative comparisons.

In the first stage, to obtain a set of robust forecasts of the evolution of TikTok, the following models were chosen:

These models were used to fit TikTok’s user base evolution data using the tools provided by the scipy toolkit in python. However, comparing the results of this process requires a more nuanced approach than goodness of fit metrics based only on discrepancies between the data and the fitting curve (e.g., Root Mean Square Error¹⁰⁵). As a model escalates in complexity, so does the risk of overfitting.¹⁰⁶ For this reason, the performance of the different models is compared using different wellestablished metrics that account for the complexity, penalizing models with a larger number of parameters. The comparison methods of choice were the adjusted R² coefficient¹⁰⁷ and the Bayesian Information Criterion (BIC).¹⁰⁸ The best model will have an adjusted R² closest to 1, and the lowest BIC value.

Using the parameters obtained from fitting the user data, the models were extended to provide projections of the number of users that the platform will have in the year 2030, effectively producing an estimate of the total size of the market. These values are used as a safety check to ensure that the newly developed model introduced in this study conforms to real user base expectations. The comparison of the different models and their forecasts can be found in Section V.

The second stage of this process involves fitting the novel dynamic model for TikTok developed in this study to user base size and average daily usage time, using the mathematical implementation of Section IV-D. By doing this, the number of potential customers N₀, the personal preferences parameters pp_j and pp_l, and the constant toa₀ are obtained, thus defining a baseline that accurately represents TikTok’s size and usage evolution under the current conditions.

Once the initial values of the controllable parameters are fixed following the procedure described in Section IV-C, an optimization function is defined. This function simultaneously minimizes the difference between the total number of users in the model (E +U +A) and the real data on the number of users, and the difference between the evolution of the time-on-app parameter (toa) and the data for the average daily time spent on the platform. The unknown parameters are automatically adjusted so that both differences are minimized. To run the model for each set of parameters, the model employs a solver for the set of differential equations equivalent to the one described in Section IV-D. The minimization algorithm chosen was the L-BFGS-B algorithm, using the implementation from the scipy package in Python.

The optimization function is used to find the best set of parameters for both choices of network effects function, i.e., Metcalfe’s law and Odlyzko-Tilly’s law. Similarly to the first stage, the model is computed again with this set of parameters, extending it to produce a forecast for the number of TikTok users in 2030. The forecasts are contrasted with the ones produced in the first stage, to determine whether the new model produces results that are compatible with the expectations set by the reference models. If this condition is fulfilled, the complete set of parameters in the baseline case is successfully defined, and the model is capable of reproducing expected evolution trends.

A. Analysis of individual indicators

First, the adoption of the service is delayed to different degrees due to the implementation of measures in different scenarios, as evidenced by the increase in users of the Explorers type. However, since the number of Explorers is still relatively small in comparison to the rest of the user groups, it is possible to argue that the delay is not very significant.

An increase in the number of Moderate Users suggests that the proposed measures are able to promote more balance and responsibility in the usage patterns of the consumers of short form video. This is especially true for Content measures, implying that Content measures may be a way of harnessing network effects for curbing excessive use and increasing wellbeing while maintaining engagement.

A significant share of the environmental and social benefits observed in the different scenarios are due to a steep decline in the proportion of Heavy Users when different measures are in place. Despite these benefits, since Heavy Users generate the most revenue, this reduction may be challenging for businesses. This may suggest the need for alternative revenue streams, and for opening the discussion on alternative business models that might eventually reward moderation (see discussion in Section VI-C).

An increment in the number of Quitters of the platform when the measures are implemented is a sign of the importance of maintaining the balance between stringent measures that deter excessive use and user retention. Some degree of churning is to be expected with any measure that may (negatively) impact user experience. The intensity of the churning, however, will depend on the strength of the measures and how the companies’ business models adapt in response.

Revenue is based on total usage of the platform, depending on the number of users and how active they are. Thus, measures that directly impact those metrics can be detrimental in terms of revenue. Content measures impact the company’s revenue the least, making them a less invasive alternative solution to the environmental and social challenges that arise from the excessive use of short-form video platforms. However, if measures from the Money or Design categories are implemented, companies may need to adapt and/or develop other monetization models.

Taxes are an effective strategy to reduce negative externalities of platforms such as TikTok, and governmental agencies should consider them as a tool in this context. Companies, on the other hand, need to plan for a future where environmental responsibilities are legal obligation.

A lower value of Experience Customization across different scenarios results from higher costs (such as data acquisition costs and taxes) and more moderate service usage. This leads to less data being collected for algorithm improvement and reduced revenue generation. Analysing the consequences of lower Experience Customization in isolation is challenging and not very productive due to its central role in the business model and the many interrelated factors. Moreover, feedback loops complicate the situation further. For instance, reduced Experience Customization can decrease usage, which in turn limits the data available for improvement, further impacting both Experience Customization and usage.

In general, the current business model of short-form video applications aims to maximize usage by enhancing the personalization of the user experience. Therefore, measures that reduce Experience Customization interfere with current practices, especially when Money and Design measures are implemented. Since content creation can increase engagement, even if it promotes moderation, Content measures are less effective in reducing Experience Customization.

Despite the challenges for businesses, the reduction in Experience Customization is particularly powerful due to decreased data usage, leading to positive environmental outcomes and potential privacy improvements. Consequently, if legislators implement measures that significantly affect this variable, short-form video businesses should adapt their business models or find alternative ways to offer a tailored experience without heavily relying on data.

The Time on App parameter faces a similar situation regarding its relationships, centrality, and potential feedback loops. Companies aim to maximize Time on App to increase revenue, while legislators may seek to minimize it to address potential environmental and mental health consequences. This highlights the importance of balancing profitability with restrictions and raises the question of whether this balance is achievable within the context of our current social and climate goals while also maintaining the existing short-form video business model. One insight for regulators is that it may be possible to reduce the negative externalities of excessive usage by implementing non-punitive design features, such as behavioural nudges.

All proposed measures result in a reduction in Environmental Impact. This is a positive message for regulators in environmental policy. Depending on the preferred approach to regulation, different effective choices can be made. However, these different choices may have varying degrees of impact. Financial restrictions (i.e., Money measures) have proven to be an effective solution, although Design measures are the most promising in this regard. This implies that regulatory efforts to address excessive screen time (e.g., ethical design, lower reliance on algorithmic recommendations) may have a ripple effect that reduces the ecological footprint of the app. Content measures, being likely more inexpensive and easier to implement, also show promising improvements, although to a lesser extent. These results also highlight the need for companies to adapt their models to prioritize sustainability in light of possible future regulations.

A similar situation is true for Mental Health Impact. In this case, however, the effect of Content measures is more pronounced, suggesting that they may be an effective approach if the objective is addressing overconsumption of short-form video. The fact that Design measures have an outstanding effect in this context highlights the importance of features such as time cues, breaking the addictive design, and changing the algorithm to reduce over-engagement. Companies should therefore prioritize user well-being to avoid the consequences of more stringent regulation, which could lead to a severe reduction in revenue.

B. Scenario impacts

A joint vision of the effects of Money measures shows that more traditional financial instruments such as increased taxes, environmental taxation or increased costs of data collection are effective in reducing environmental and mental health impacts of short-form video applications, promoting a more moderate consumption of these services. In the same way as the other measures, they come at the cost of reducing revenue for the business that offers those services. However, this type of regulation might not provide a definitive solution in some cases, and alternative types of measures should be considered.

One option are Content measures, which offer a less intrusive (although somewhat less effective) approach to addressing issues such as environmental and mental health impacts. These measures directly exploit the power of network effects, which on the one hand lead to the fast growth of social platforms, but on the other can be used to disseminate useful information. These results show that behaviour can be influenced from within the app, without drastic intervention, and still have an effect. Avoiding very stringent regulation may lead to reduced pushback from companies and lobbies, increasing the chances of a successful intervention. Another advantage of these measures is that they can be easily scaled across regions and platforms, while remaining relatively low-cost. For the reasons described, Content measures are a relatively common topic of research in the area of short-form video as a tool for sustainability, both on the social^14,30–32 and environmental^{33,34,54–56} dimensions.

Design measures likely show the most promising results as a stand-alone set of measures. Not only do they significantly reduce mental health and environmental impacts, surpassing the results from Money measures, but they also manage to preserve similar or even higher revenue. This is because these measures specifically address the addictive features of these applications, aligning with research on the influence of digital interfaces on people’s behaviour.¹⁰⁹ Being a non-punitive category of measures, these results reveal an alternative approach for mitigating the negative externalities of short-form video platforms.

A key takeaway from this scenario is that, although implementing Design measures could initially be done from a public health concern perspective, it would also create a ripple effect, leading to considerable reductions in the environmental impact of shortform video. Thus, these results provide evidence of an indirect link between UI/UX design and environmental sustainability, suggesting the need for collaboration between UI/UX experts and policymakers.

The last piece of the analysis is a combination of All scenarios, where the measures of each individual scenario are applied to a more moderate extent, directly focusing on the balance between effectiveness and feasibility. Since no individual measure is applied to an extreme degree, this scenario offers evidence of the benefits of a complex, balanced, multi-pronged approach. In this scenario, the strongest reductions in environmental and mental health impacts are observed, whilst still maintaining some level of engagement.

For policymakers, this signifies that there are clear advantages when financial, social, and design measures are jointly implemented, compared to any one of them in isolation. This is particularly useful in cases of scepticism about strong financial restrictions and taxes. Although these are effective methods for addressing the issues discussed in this research, they can be substantially relaxed if implemented in conjunction with Design reforms and Content tools.

This policy framework thus proposed should serve as a starting point for researchers to study the interactions between financial, social, and design measures, and their effect on user behaviour.

A key takeaway for companies is a call to action to explore new business models that better align with our sustainability goals, as future regulations and measures may result in significant financial impacts. Adapting to these changes could mean the difference between reaping the benefits of alternative, more sustainable, long-term oriented business practices, or eventually being unable to continue operations, losing the battle against more sustainable competitors.

C. Alternative business models

As a final note for the discussion, and as a starting point for further research, one can consider several alternative business models that could be studied for viability in the short-form video industry. These models are engineered to facilitate the adaptation of short-form video businesses to possible new sustainability regulations, and help them thrive in a more responsible way.

A Premium Subscription business model might be hard to conceive in the current state of the industry, since the main appeal of short-form video applications is the fact that they are free to use. However, this model is not only more profitable in certain conditions,^73,110 but has also proven to be a more sustainable alternative due to the decrease in traffic and data processing required by advertising.⁹¹ A Premium model could have the additional benefit of increasing average content quality, as it would reduce the need for advertisements displayed in the form of content, while also increasing content creators’ revenues. This could potentially reduce their need for paid brand partnerships, giving them the opportunity to prioritize content creation over product promotion. Nonetheless, since an abrupt transition to a Premium model could be unreasonable for many companies in the short-form video industry, there is a need for exploration of other business models.

A Freemium model¹¹¹ is a compromise between the current free model with ads and a more sustainable Premium model. Although the environmental benefits may not be as large as with the Premium-only model, they could be significant compared to the status quo. This model implies a change of incentives. Rather than maximizing ad visualization, the company’s objective becomes to “funnel”¹¹¹ as many users as possible to the premium category. This change in incentives would result in a degradation of the free-user experience, either by removing or limiting access to certain features, including premium-only features, or by modifying the advertisement strategy to incentivize users to purchase the premium option. Such degradation could reduce engagement measures for free users, such as time spent on the app, leading to some of the positive environmental and social impacts described in this research. Moreover, since the premium option would likely have lower impacts too,^73,91 all user groups in this case would participate in these positive impacts.

A variation of the Freemium model is the Tiered Engagement model. This model sets a limit on how much content a free user is allowed to consume on the app, following recommendations on time-limits from research.^13,39 The free version may include some ads, but since it does not rely exclusively on those, the ads-to-content ratio can be smaller than it currently is. In addition to the free version, there could be several premium tiers of consumption. For instance, one tier would allow a user to watch five times as many videos as a free user, and the next tier could be an unlimited ad-free service. The tier architecture can be adapted to different cases. Since the time spent on the app is limited by design for certain user groups, this model has a moderation-promoting feature integrated into its core. Moreover, the limited engagement time also reduces the data gathered by the business, affecting further engagement improvements. All of these factors contribute to lowering the impacts studied in this research. This model could be a viable monetization strategy, as it is very similar to the one currently used by the dating app Tinder.¹¹²

One alternative model that focuses on personal data use and ownership is what we call Data-Based Freemium. This model would be a response to measures such as data collection charges or a data tax. Here, users have the option to opt in to data sharing in exchange for a more premium-like service. User data is the primary source for improving both video recommendations and ad recommendations, being a crucial resource for companies that provide short-form video content. Any user may choose one of two options: 1) Not sharing their data and receiving less targeted content, both for ads and videos, or 2) Sharing their data and receiving fewer ads and/or gaining access to special features. Because non-sharing users receive a less personalized experience (lower value of Experience Customization variable), their engagement will be lower, improving environmental and mental health indicators. Moreover, because data availability would be lower than in the current business model, the level of personalization for sharing users would also be lower, affecting engagement. Users who do get a personalized experience by sharing their data are subject to some of the benefits described for the Premium Subscription model, further reducing environmental and mental health impacts.

Finally, all the previously described models could benefit from a conscious choice of partnerships. Since several of the suggested models still partially rely on advertising, choosing advertisers that align with our sustainability goals could add value to the efforts of the company for mitigating negative impacts. Some possibilities are 1) charging advertisers of sustainability-oriented projects smaller fees, 2) displaying sustainability-oriented ads with a higher relative frequency, 3) using corporate sponsorships to reduce overall advertising, or 4) using brand sponsorships to fund content-driven initiatives (Content Measures). These suggestions can have a positive contribution even with the current business model, although they may not be enough if future regulations incentivize a true change in the business model for short-form video companies. For this reason, we suggest they are explored as an addition to possible alternative approaches to monetization of short-form video content.

A. Limitations and future work

First, the evidence provided is based on simulation. Despite the model being carefully developed using scientific principles, real data, and well-known techniques for modelling digital applications, simulation evidence is always less reliable than direct evidence. Future research should explore how already existing measures have impacted short-form video and use those insights to fine-tune the models developed in this work.

To improve our approach to policy modelling, it is imperative to seek collaboration with experts and government officials, whose feedback will be extremely valuable for developing more complex and nuanced scenarios. This collaboration would result in important insights for future decision-making in sustainability and technological policy.

The models presented in this work can be extended to incorporate competition between several short-form video companies, which may have different business models. Some business models that could be explored and compared were presented in Section VI-C. Another open research line where these models could be useful would be the comparison of long versus short video formats and their sustainability impacts, alongside measures to promote the most sustainable business alternatives.

This research only used open access data, limiting the type of information that could be used for modelling a real service. More accurate but privately-held data from companies could be used to achieve higher precision in future studies, although this would certainly limit reproducibility. This study focuses specifically on data about the platform TikTok, whose operations could be severely compromised by its political situation in the US regarding privacy concerns.²³ Although TikTok is still active at the time of the writing of this article, due to the uncertainty about its future, it is crucial to emphasize that our core findings can be extrapolated to other platforms with a similar business model, since the modelling strategy is data-independent. Applying it to a new service would require a new fit and normalization of some of the model’s parameters.

One potential future research line is developing models that incorporate different user segments, and analyse different impacts depending on the characteristics of individual user groups. This approach would allow to factor in nuanced differences such as gender-specific impacts, as the effects of excessive short-form video consumption have been shown to differ based on the user’s gender.³⁹

Finally, this work favours incentivising future interdisciplinary research, with experts from behavioural science, environmental science, public health, and economics. An interdisciplinary perspective is vital for developing a nuanced and in-depth comprehension of the complex task of making digital services and ecosystems more sustainable.