In this episode of The iGaming Show, we talk about how AI can be leveraged to enhance responsible gambling initiatives.

Guest: Rasmus Kjaergaard, CEO at Mindway AI

The iGaming Show, presented by Paramount Commerce, is a podcast that will analyse gaming industry trends with experts from various gaming organizations.

Please like, comment, and share this video. Also, stay up to date with our content by subscribing to our YouTube channel.

Full episode transcript:

Varad Mehta: Hello everyone, and welcome to the 19th episode of The iGaming Show presented by Paramount Commerce. I’m your host, Varad Mehta, and in this podcast, we analyse gaming industry trends with experts from various gaming organisations. In today’s episode, we will look at how AI is being used for responsible gambling with Rasmus Kjaergaard, the CEO at Mindway AI. So without further ado, Let’s get the show rolling. Rasmus, how we always begin this podcast is by asking our guests a few fun questions. I have a couple lined up for you. My first one is that, how much do you love the Danish football club known as AGF? And what does KSDH mean?

Rasmus Kjaergaard: Oh, that’s a fun one and brilliant research there. So AGF is my favourite hometown or the club from the city where I am born. I’m originally born as a Aarhusian meaning I’m coming from Aarhus, which is the second largest city in Denmark. The A stands actually for Aarhus FC, just in Danish words. So this has been my soccer team, and I had my first home game at the Aarhus Stadium back in November 1987. So that has been a while. And just this Sunday, they drew against FC Copenhagen. And that actually led to that Aarhus is now leading the Danish Super League leaderboard or table after or 15 rounds. And that has been, let’s say, three decades ago since that happened last time at the 15 rounds. So we are excited and we hope that that will last for the rest of the season, which is we are almost halfway through approaching the winter break. KSDH stands for “come on your whites” and Aarhus plays in white shirts and naval blue jerseys. So that’s basically the translation in English.

VM: I love that. There’s some football love happening here.

RK: Quite a lot. Yes, definitely. I’m a huge, huge soccer fan.

VM: I love that. My second question for you, Rasmus, is what two Bruce Springsteen songs would you recommend to someone who hasn’t heard him before?

RK: Well, that’s also somewhat a difficult one. I would say I like a lot, actually, it’s my favourite song, the song called Badlands from the album, Darkness on the Edge of Town, which is such an energetic tune, and also with lyrics that I can definitely relate to. Also, it’s, as I recall it, the second most live-played track he has, only surpassed by Born to Run, which is his signature number. As a huge Bruce Springsteen fan, almost as huge, if not huge, I don’t know. I can’t assess the difference between being a Bruce Springsteen or soccer fan, the level of fanship there. But a huge Bruce Springsteen fan as well, I would say Born to Run as the second one. But the catalogue is so big there that I could have mentioned 10, 15, 20 other songs as well.

VM: Love that. Now, going to the topic of discussion today, we’re talking about AI and how it’s going to collaborate with responsible gaming. My first question for you would be that currently, how have you seen AI being used for responsible gambling, or is it even being used right now?

RK: Oh yeah, quite a lot. We are one of the ones providing such solutions. There are also other ones, and also to a wide extent, also in-house developed solutions by operators themselves. When I started, we had no customers, no solutions developed. Now we have another position in this space where we provide solutions to gambling operators currently in 38 countries and in four continents, including Canada, Ontario currently, and also USA, where we have our game scanner tracking the gambling behaviour of more than 8 million active players per month. In that definition, active players means real revenue-generating players, so excluding free spins, bonus bets, or whatever it’s called. We provide this solution for gambling operators of all sizes. Our smallest operator has between 2,000 and 3,000 active players locally here in Denmark. The biggest ones between 3 and 4 million players per month using our software in 30-something countries. We work with the likes of Entain, Evoke, Flutter, Tabcorp in Australia, but also a longer list of medium-sized regional and local one country only. We support all game types, meaning both sports betting-related online casino games, slot machines, lotteries online. Moving into land-based as well lately with a couple of current negotiations where we want to, together with the land-based operator, to utilise the fact that we, onwards, are more and more able to identify the player via either carded loyalty playing or mandatory player cards in some jurisdictions. Together with the opportunity to get data from the gambling platform, also land-based, some of the current platforms there and moving into also looking into how can data be generated with more analogue table games, for instance, the typical card and wheel tables. We have this opportunity to do the same analysis in land-based settings as we do online. That will also accommodate the operators where they have customers using both land-based casinos and online casinos as well with the same operator. Definitely, to a very large extent, these solutions are already being used for the purpose of responsible gambling.

VM: You just mentioned, I’d love for you to go more in detail, you mentioned the game scanner. Could you provide more information about what Mindway has done with AI to enhance responsible gambling tools, initiatives? Have there been any challenges associated with that?

RK: Yes. We have taken very early days, actually, that’s part of the whole key of what we have built our company on this innovative approach where we combine human expert assessment with the use of AI or algorithms. In the creation of this scientific and technological foundation, our co-founder, Professor Kim Mouridsenfrom Aarhus University, in his research, found that most solutions for finding at-risk or problem gamblers or more widely defined as for tracking the full picture of the gambling behaviour on the individual customer. We saw a space where there was other solutions created, other approaches created and built based on either self-exclusion, self-assessment, or changes in behaviour. And what we found, or what Kim found in particular, was that neither of those targets for the algorithm works sufficiently clear because self-exclusion is not used by all problem gamblers. Training the algorithm upon that will give an unclear outcome. The same with self-assessment, where typical questionnaires, actually, we created a self-test as a cardgame to cope with that, but those questionnaires are biased due to the fact that questions are not answered completely truthfully as they are sensitive questions. Also asking a problem gambler on how much they gamble during a week is likely not answered truthfully. Finally, the changes in behaviour is not a clear target due to the fact that a high-risk behaviour can be contained without a change for a long time, where such a behaviour will not be flagged in the system based on that. Also the whole in parallel challenge is that less, let’s say, holistic approach where you try to define a risk group on something that is completely human individual, like a gambling behaviour, which is a human individual behaviour. It’s not necessarily, in our view, the best way to try to find at-risk and problem gamblers. To accommodate all this and to strive for the objective to use AI for this purpose of finding at-risk and problem gamblers. We found that there was a need for a stronger target, a stronger approach where we could meet what we wanted in terms of clarifying the individual gamblers, individual gambling behaviour, but still use AI for the purpose to clarify the full picture, not only the monetary spent or the time spent, but the full picture also with the behavioural markets, and even much more important actually, in my view, the whole dynamics between all of these different behavioural factors. And that is actually, you would say, the secret sauce what we have also US patented nowadays, EU patent pending, where we combine our algorithms with, I mentioned it a couple of times, human experts. And in our team of experts or expert panel is typically a psychologist working with RG treatment. And also now, lately, we have a partnership with Epic Global Solutions, where we have added lived experience capabilities from them into the expert panel. And that gives us the opportunity to, today, okay, if you were signing up as an operator, we provide our pre-trained model first and foremost, because that has a lot of different trainings aggregated into it, still trained by experts, but pre-trained. So we have now quite a sophisticated out-of-the-box model. But if the operator for different reasons could be because now you’re based in Ontario, but we see movements in other Canadian states, Alberta, for instance – if you have had an opening there, you would potentially want to tune your model towards specific local cultural stuff. Or if you were in Ontario and wanted to have deploy a solution in Macau, risk appetite is significantly different I would anticipate. I have not been to Macau, but I have been to Ontario, and I’m pretty sure that’s a fair statement. For such purposes or for the purpose of the operator wanting to be able to say, this is a solution that utilises this significant approach, but tuned on my own data, then we will build a specific model based on specific experts to that purpose, briefing the experts on cultural differences and could be also a specificities related to if there were specific local game types, for instance, you have craps in the US, you have pelota, which is a sport in Catalonia, in Spain, as I recall it. If you had a lot of such games which we haven’t seen that much of in the model, we can factor that in. That gives us a flexibility to try our best to create the highest possible accuracy in the detection, the tracking of the clarification of the full picture of the gambling behaviour. But it’s important for me to underline also that operators are not, let’s say, in a situation where they affect how we build the model. Actually, most of our, if not all of our operator customers like that because that’s part what they’re also built to be able to say, We can’t affect it, if that makes sense.

VM: Continuing on that thought, then, you just mentioned something so interesting that you built these diverse AI models. But what role do you think then those models can play in tailoring interventions for at-risk players?

RK: A huge role. Our customers use the tracking of the gambling behaviour to be an embedded part of their intervention strategies or their intervention programme or their communication programme or whatever they call it. Depending on the jurisdiction they’re operating in, you have going from very high obligations to average obligations to no obligations, and even low obligations to even no obligations in some jurisdictions. Let’s say we are operational in all sorts of such graduation of intervention obligations. Definitely, following the customer with the right level of intervention or communication at the right point in time is very important. I think there’s different layers in this. First and foremost, the whole translation and interpretation on what the model actually has an outcome and how you use it. The next level layer is the fact that with a high efficacy, where we have tested and validated our software with GLI, and they found an 87% of all cases, we’re able to find problem gamblers compared to a human psychologist. On both sides of such an accuracy, you actually have opportunities also as an operator. Obviously, we will flag more of the false negative, so we will find more and more sophisticated full gambling behaviour detection. Of course, short-term potentially an issue for the operator. However, for the management of the operators, I would say as a CEO myself, I would any day prefer to know my risk level compared to not knowing, because then I know how I can navigate and mitigate. Maybe not all C-level persons of gambling operators would agree with me on that. However, you don’t find yourself, they are the ones who can cost a huge loss of reputation or media stories or even in some jurisdictions, quite huge fines. On the other side of the accuracy, the false positives. If we can, and we do that, decrease the number of those, actually, that’s a business opportunity for operators because they are some of the best customers. We don’t have to bother as much or at all, which were flagged as high risk in a less accurate, less sophisticated solution. That’s the other layer. Then obviously, I would say as a conclusion, given the right intervention at the right time is crucial because the more we can nudge the player to a more sustainable behaviour, more individualised earlier based on the player’s individual gambling behaviour, which is exactly what we can provide, is also a potential business opportunity for the operator to utilise the high cost of acquiring new customers. But also once a player has spun out of control, high risk, even potentially play break, close of account,self-exclusion, then that’s more or less a customer gone forever. And also if a customer were spinning into quite a big issue personally,  obviously, that’s always this tricky conundrum because operators would be reluctant to sometimes go into, and that could be one of the challenges we have encountered, into such conversations because they are afraid of losing their best customers. But basically, if we could decrease their best customers and make them stay around for longer, because one of the confirmations we have from our psychologist expert says that once customers are spun out of up to a high-risk level and likely also have developed an addiction, we do not diagnose, so we assess risk. But we also know for a fact that there’s a high correlation betweenthe high-risk players we can flag in the tracking system or we assess as such. There’s high correlation to people who in the treatment space, turn out as addict, unfortunately. And once that line has been crossed, it’s not really possible to decrease the level or go back long term. So for both the player and the operator’s sake, we would hope that utilising what we provide to, in general, lower the risk-level is what we want to achieve.

VM: I’d like to tie two questions into my last one. Obviously, I’d love to know what is the future of AI in responsible gambling and what your team is doing, but also talking about how privacy will come into that conversation right now. This is such a new topic when we talk about AI, but when you take it with responsible gambling and having the approach you’re talking about, what do you think privacy will play as time progresses and what is your team going to work on in the future?

RK: Well, yeah. Let’s start with the privacy part. So if I revert to, I was saying the vast majority should entertain themselves, not getting bothered, but the few percentage of customer databases who show signs of at-risk and problem gambling should have the proper help, basically. That’s what we provide the whole idea in our position. I very often also refer to such systems as lane-keeping assistance. We provide a lane-keeping assistant to the operator where they can reach out to the customer showing signs of deviating from the road, basically. You would never, ever dream of buying a new car without a lane-keeping assistant nowadays, nor would you buy or you are even not allowed to enter an aeroplane by airline or aeroplane authorities or aviation authorities without having all sorts of safety systems. But basically, we allow this in many jurisdictions, not least in jurisdictions in North America. So that’s on the other side of the privacy. I do think that such safety system should be in place, not that they should be from us. Obviously, that would be nice. But in general, I think that when you provide such entertainment, safety systems have to follow through. But when it comes to privacy, we actually have built that into what we provide. We are an E-danish-based and registered company, meaning we are part of the European Union. We are definitely  a first-hand experience on GDPR and comply with similar regulation in North America as well. When we provide the software in terms of privacy and, let’s say, sharing of these data, they are pseudonymized. We do not know who the player is. The operator knows who the player is, but they know who the player is anyway because they have to due to KYC or AML or other type of purposes. I think when you talk about privacy, when it comes to seeing from the consumer perspective, I also think that is tied – when we speak about AI-based solutions, there’s a couple of different layers in that. First and foremost, AI is not just AI. There’s a lot of different types of AI. The whole reason why we also chose to have human-supervised algorithms was to avoid black box algorithms, to avoid these algorithms going rogue. Our solutions are trained by human experts because problem gambling is a sensitive matter, and we cannot in itself rely on computers that develop themselves towards something that is not in alignment of what human experts in this sensitive field is doing. Then we can add sophisticated AI elsewhere. But for the key part, in my view, of the detection, we still need to have a human supervision from such experts, in my view. But that also gives us the opportunity to provide the software in an explainable AI-based way. Everybody who uses our software knows exactly what the outcome is, knows exactly what the result is, how to interpret this. This is in itself also ongoingly assessed by industry peers, by our experts, open to take into consideration new research, new standardisation. We are part of the Markers of Harms Standardisation Project in EU as well. Then on top of this, as a software provider, we also do a lot to build higher and higher, let’s say, tech credibility or tech sophistication. We provide the software in the operator’s own environment so the operator can control themselves, who they will share the data with. We also have a hosted version. We are an AWS partner, and we work on more significant enhancements on such tech sites to ensure the privacy in the best way possible.

VM: Okay. And talking about the future of Mindway AI and what you’re going to do in the future, I’d love to know what exciting things you’re working on.

RK: Yeah. Well, it is exciting times. We have a lot of ideas. We have, as I mentioned, grown rapidly since we started, basically five, six years ago. We have come to a position where we can also drive new initiatives on our own side, I would say. Obviously, more sophisticated functionality and solutions around how to communicate more effectively and individually with players is a key. Obviously, it is a fact that specific player groups, and particularly vulnerable groups like the young boys versus older guys like myself have to get presented such messages in different ways. I can be reached by an email most easily, I would say. If it is go wide, maybe a WhatsApp or a text message or maybe even on Messenger. My son, not at all, he uses other channels. For the Gen Z, obviously also not only other channels, but also other measures to communicate. Definitely, that’s something we are exploring. And developing functionality around also, obviously, as an AI-based company, large language models are obvious also to see how can we integrate the new disciplines of AI into what we provide around the core of what I mentioned before in terms of maintaining the detection human supervised. But of course, everything else where we can add in faster decision-making support for better detection and intervention, we will do so. On the wider scale, I think it’s important that we take part of, and we have done so both in terms of the markers of harm project and in other initiatives, onwards to drive enhancement of standards on this area. As said before, in most jurisdictions, you don’t even have to have such safety systems. In the jurisdictions where you have to have such safety systems, the standards are very different and basically locally founded. So it goes from a very instructive and in some operators’ views, hard regulation up here in Northern Europe, Denmark, Sweden, the Netherlands, UK, Germany to some extent, to elsewhere. And the idea of enhancing the standards on how such models are built to increase the efficacy is definitely something we will push and drive for, as well as how to standardise how we detect at-risk and problem gambling and the categorization of such.

VM: That’s fantastic. Rasmus, thank you so much for sparing your time and providing all these amazing insights into what you and your team are working on. When it comes to responsible gambling, having these conversations is so important, and finding innovation in that specific category is so unique. And thank you so much for telling us all about the amazing work that’s happening on your end. I hope whoever listens to this can really benefit from what Rasmus has shared with us today. And also all the Bruce Springsteen songs that he has recommended to us. So thank you again, Rasmus, for joining us today and providing your expertise.

RK: Thank you very much. Thank you for having me.

VM: Whether it is enhancing player safety or detecting risky behaviour, AI is going to play an important role in creating unique responsible gambling initiatives and tools. I want to thank Rasmus Kjaergaard, the CEO at Mindway AI, for joining us today and providing his expertise. If any questions for us or Rasmus, please do comment them down below. Please don’t forget to like and share this episode and subscribe to our YouTube channel. For the episode transcript and more amazing content, please visit: paramountcommerce.com Thank you so much for tuning into The iGaming Show presented by Paramount Commerce. I’m your host, Varad Mehta and until next time, keep gaming.