How Our Chatbot Helped People Lose 49% More Weight Than a Control Group
This summer, our health & fitness chatbot, Jolt.ai, helped people lose 49% more weight and 227% more body fat than a control group.
While my team and I were were thrilled to see these figures, we weren’t exactly taken aback. We’d devoted significant time and energy to the behavioral change components of Jolt.ai, and we understood its utility firsthand.
Before I go into details on these impressive results, let me breakdown how we built a behavior change & reinforcement bot that actually works.
Understanding Habit Formation
Habit formation is a BIG subject and I won’t go into too much detail here, but the gist is this: Behavior change is hard. Anyone who’s ever made a New Year’s resolution can attest to this.
For one, we’re often motivated by negative feelings like guilt, fear, or regret. One analysis of 129 behavior-change strategies, however, found that the least effective strategies are those prompted by negative emotions.
Guilt and regret aren’t great engines for change.
Top Articles on How Businesses are using Bots:
Plus there’s our biology.
Enjoyable behaviors prompt the release of dopamine, a chemical that can make your brain associate pleasure with a behavior. The more you do the behavior, the more the association is reinforced. A habit forms.
Of course, our behavior also isn’t fixed. We can change. And we do.
We’ve all, at one point or another, broken a bad habit or found a way of keeping up a good one.
So when my team and I at Rock My World Media were building Jolt.ai, we kept this fact in the forefront of our minds — that positive changes really can be made and maintained.
A key to this is recognizing and understanding what motivates — and demotivates — us.
(PS — If you’re looking for a great book on habit formation, consider “The Power of Habit.” I loved it!)
The world of motivation
For those who aren’t familiar, there are two types of motivation: intrinsic and extrinsic.
Intrinsic motivation is internal. Whether it’s for improved health, more energy, or any other sort of physical or mental benefit, intrinsically-motivated exercisers do it because they want to. Outside forces aren’t nudging them.
Extrinsic motivation is all about nudges. A weight loss contest at your office, gentle (or not-so-gentle) prodding from your doctor or spouse, a gym raffling off a jetski to its members, and so on.
If anything other than you is what’s motivating you to do something, it’s an extrinsic factor.
Extrinsic motivation alone typically isn’t effective for the long-term.
What happens, for example, if you’ve lost weight but eventually plateau? If your doctor’s now satisfied with your blood pressure numbers? Or, critically, if you don’t win that jetski?
The incentive won’t be there. And because the desire for the incentive was greater than the desire to be in shape, there’s a solid chance that your commitment will wane. The incentive was your fuel.
Intrinsic motivation, meanwhile, doesn’t cut it for everyone. Some people do have the internal drive to, say, make it to the gym at 5:30am every day. And they want to. But they aren’t exactly the norm.
So when we were designing Jolt.ai, we came to a straightforward conclusion: it would need to 1) help foster intrinsic motivation AND 2) provide extrinsic motivation.
These factors in action
To help illustrate this, let me give you a quick walk-through of Jolt.ai.
First, Jolt.ai suggests a movement goal for the week. The goal is determined, in part, by a user’s exercise frequency, fitness aim, and weight. (After the first week, our algorithm also factors in historical movement data.)
The user can then raise, lower, or confirm the suggested goal. Because autonomy, the feeling that you’re in control, is a critical aspect of motivation, Jolt.ai doesn’t make any choices for the user.
Jolt.ai, instead, supports.
From there, users receive daily activity updates, can compete on a leaderboard, earn “JoltCoins,” and encourage others, among other features.
These features aren’t accidental. They’re designed to help users build a positive feedback loop — develop a routine they enjoy — and to be encouraged along the way.
But the nuances involved didn’t come overnight.
Recognition, for example.
During our first round of beta testing, some of the key feedback we received involved users wanting to be recognized for their efforts.
If they had exercised a lot in a given day, or had been on a roll, hitting goals for multiple weeks, they wanted someone (yep, even a bot) to recognize as much.
We incorporated “Achievements” — kudos, of sorts, for breaking various personal records.
These Achievements really do feel good to receive.
I’ve used Jolt.ai for about a year now and exercise frequently — and I too, even as someone so close to the product, feel a sense of pride when Jolt.ai recognizes that I’ve set a personal record.
Who doesn’t like to know when they’re doing well?
Next, we began tailoring content around frequently-logged activities.
When a user logs an activity, say, “HIIT for 30 minutes,” Jolt.ai confirms the input and awards them a certain number of “MovePoints.” (MovePoints are Jolt.ai’s unit of measure — the more intense the activity, and the longer one does it, the more MovePoints one earns.)
We then deliver content tailored around that input. The content might be a joke, or a high-five for completing a hard workout.
The bottom line is that Jolt.ai recognizes their specific efforts.
These messages have received particularly high rates of positive responses (things like, “Thanks!” “haha,” “lol,” etc.). During feedback sessions, moreover, users have shared that they’re more likely to record and track activities when they receive this kind of content. It feels special.
Positive recognition, however small, is an important agent for behavior reinforcement.
Heading to the CoinStore
In recent months we also implemented perhaps the quintessential incentive: rewards.
Heart rate monitors, Starbucks gift cards, supplements, bottles of wine. Every day of the week, we provide our users with these rewards, and more.
It’s a simple process. In addition earning to MovePoints, users earn JoltCoins, which can be redeemed for items in Jolt.ai’s CoinStore.
Before rolling out this feature, we surveyed 250 users to gauge enthusiasm: Do you agree or disagree with the following statement — “I’d be even more motivated to earn more MovePoints if they could help me earn rewards and prizes.”
85% of respondents “strongly” or “somewhat” agreed. (The remainder were neutral.)
After we received this data, one of my teammates provided some thorough analysis: “Yeah. People like free stuff.”
He’s right. People do. But free stuff — any incentive — should only be a component of a behavior change & reinforcement strategy, not the strategy itself.
In a now-classic work, “Punished by Rewards,” the social scientist Alfie Kohn argues that rewards can be demotivating for the intrinsically-motivated.
If an incentive is less meaningful than one’s own sense of satisfaction, the incentive might turn out to be counterproductive. It can mitigate inspiration.
Internal drive, whenever we have it, is a potent force, and it doesn’t like to be tinkered with.
For my teammates and I, then, it’s been crucial to offer rewards as great frosting. Not the whole cake.
…And the results?
As I mentioned, this summer a group of Jolt.ai users lost significantly more weight and body fat than their non-Jolt counterparts.
Here are the details.
One of our partners hosts an eight-week fitness challenge for its members. As part of the most recent challenge, all of the participants were offered Jolt.ai.
Those who used Jolt.ai would also be entered to win a spa day — the more MovePoints they earned, the more “tickets” they’d earn for the spa raffle.
About a third of the challenge participants signed up for Jolt.ai and used it throughout the eight weeks.
Following the challenge, we had two clear cut samples: those who used Jolt.ai, and those who didn’t.
Correlation, of course, does not equal causation, and this was not a large, peer-reviewed study.
But the data were compelling: Jolt.ai users, on average, lost 49% more weight (4.9lb. vs. 3.3lb.) and 227% more body fat (4.48% vs. 1.37%).
In the end, the participants’ debriefs read like lists of behavior change and reinforcement strategies: “Receiving the activity updates reminded me to keep up the workouts,” “I like spa days…” “I was trying to be on top of the leaderboard,” “I kept wanting to beat my numbers from the previous week…”
The interplay of motivational strategies had broad and powerful appeal.
While Jolt.ai clearly has direct benefits for the fitness industry, the potential applications of this and similar technologies reaches far further.
For example, researchers at the Fox Chase Cancer Center in Philadelphia recently presented findings on how chemotherapy patients who received automated daily text messages from their healthcare providers reported lower overall levels of distress and a higher quality of life during treatment.
Automated behavioral change and reinforcement technologies will play increasingly important and diverse roles in our lives.
Meanwhile, my team and I are going to continue to iterate on Jolt.ai, and ensure that it’s as effective a tool as possible.
Although we’ll never be “finished,” we know that we’re moving in the right direction.
To see for yourself how Jolt.ai works, just give a tap here.
Thanks for reading about our efforts to date, and if you have any questions or comments, please add them below!