Some users of popular fitness and calorie counting apps experience shame, disappointment and demotivation, potentially undermining their health and wellbeing, according to a new study led by researchers at UCL (University College London) and Loughborough University.
The study, published in the British Journal of Health Psychology, looked at 58,881 posts on Twitter (i.e., posted prior to the platform becoming X) relating to five popular fitness apps*.
The research team used AI models to filter out 13,799 posts judged to contain negative sentiment and then to group these posts into broad themes or topics.
They found users expressing shame at having to log "unhealthy" foods, irritation at notifications to log calories or to keep sugar consumption low, and disappointment at slow progress towards algorithm-generated targets. In some cases, these experiences led to demotivation, with users seemingly giving up on their goals, which is the opposite of what they are intended to be used for.
The researchers called for fitness apps to move away from "rigid" calorie counting and exercise quantification to a more holistic approach focusing on wellbeing.
Senior author Dr Paulina Bondaronek (UCL Institute of Health Informatics) said: "Few studies have looked at the potential detrimental effects of these apps. Social media provides a huge amount of data that could help us understand these effects. By using AI, we were able to analyse this data more quickly.
"In these posts, we found a lot of blame and shame, with people feeling they were not doing as well as they should be. These emotional effects may end up harming people's motivation and their health.
"Instead of very narrow, rigid measures of success relating to amount of weight lost, health apps should prioritise overall wellbeing and focus on intrinsic motivation – i.e., the inherent enjoyment or satisfaction in activities.
"Many of these apps also ask users to do tasks individually. This misses out the great potential of social connectedness for improving our health and happiness."
Dr Bondaronek added: "Self-monitoring and action planning are powerful behaviour change techniques. But we over-use them. We need to learn to be kinder to ourselves. We are good at blaming and shaming because we think it will help us to do better but actually it has the opposite effect.
"It is important to note, too, that we only looked at negative posts, so we cannot assess the overall effect of these apps in terms of our wellbeing. The apps may have a negative side, but they likely also provide benefits to many people."
In their paper, the researchers grouped negative posts into several themes, chosen from a larger group of themes suggested by AI. These included frustration at the complexity of trying to track calories, technical faults and challenges leading to data loss, and the emotional impact of logging daily activity.
In terms of the complexity of tracking calories, the apps lacked the personalisation and tailoring needed to be fully accurate, with one user highlighting that they were unable to log breastfeeding within the app – an activity that can expend significant calories.
The researchers noted users' surprise regarding the apps' calorie recommendations, with one individual reporting being told that they needed to consume "−700 (negative 700) calories a day" to reach their goal. Another user said: "If you allow [MyFitnessPal] to prescribe your calories you'll end up with a deficit that's unachievable, unsustainable and very unhealthy. You could also starve to death ...."
Goals set by fitness apps were not based on public health recommendations (e.g. NHS recommendations for daily calorie intake), but were dictated by the user's weight goals, which could lead to unrealistic or unsafe recommendations being given.
In terms of the emotional impact of logging daily activity, the researchers observed that some users felt "pestered" by app notifications, and that the difficulty of sticking to rigid goals led to loss of motivation.
They wrote: "Where individuals faced difficulties in keeping within the targets set by the app (e.g. losing a 'streak' or not meeting their daily goal) this appeared to contribute to avoidant behaviours ('do NOT put Percy pigs into MyFitnessPal') or complete disengagement ('back to eating lotus biscoff spread out of jar')."
The team found that some of the negative sentiment expressed in the posts were in line with previous research finding that measuring an activity might decrease the enjoyment of it. They cited a Strava user who had achieved their personal best for a half marathon but, in their Twitter post, had focused on the disappointment of this time not being recorded on the app.
Another user, they said, "expressed an intention to go to the gym in response to not meeting their calorie intake goals, describing themselves as 'miserably' stuffed, indicating that their decision to exercise was driven by negative emotions and the external prompt of the app's feedback", rather than by the inherent enjoyment of the exercise.
Co-author Dr Lucy Porter, from the UCL Division of Psychology & Language Sciences, said: "When we send behaviour change tools out into the world, it's so important that we check if they have any unintended consequences. Listening to users' reports on social media has shown that fitness apps can sometimes leave users feeling demoralised and ready to give up - which is the exact opposite of what these tools are supposed to do!
"We know from previous research that feeling ashamed and miserable about yourself is not going to support healthy, long-term behaviour change - what we need to know now is how pervasive these effects on morale and emotional wellbeing are, and whether there is anything that can be done to adapt fitness apps so that they better meet people's needs."
Co-author Trisevgeni Papakonstantinou, also from UCL Psychology & Language Sciences, said: "Unsupervised machine learning (or we can say AI) lets us analyse insights from real-world social media data that would otherwise be inaccessible. By using these tools, we can turn naturalistic data into actionable insights for public benefit - and do so at scale, and at no cost."
The study involved researchers at UCL, the University of Westminster, and Loughborough University. The lead author is funded by Wellcome to investigate large-scale textual data analysis using AI.
*The research team chose the five fitness apps that had the highest revenue. They were: MyFitnessPal, Strava, WW (formerly Weight Watchers), Workouts by Muscle Booster, Fitness Coach & Diet and FitCoach. However, most of the posts (8,464) related to MyFitnessPal, with Strava and WW also accounting for a large portion of the posts (2,264 and 2,902 respectively).
