According to South Africa’s National Gambling Board’s annual report, South Africans spent R 1.5 trillion on gambling sites over the past year. Mostly on betting sites. Why is gambling so addictive? It has to do with the how people are conditioned to behave and how the brain’s reward system works.
A quick segway into behavioural and neuroscience can explain this.
People learn how to behave largely through reinforcement, which can be a positive or negative stimulus. There are a variety of forms of reinforcement but partial reinforcement is mostly associated with addictive behaviours such as gambling. Here, positive reinforcements follow a variable ratio schedule by not following every response with a reinforcement. Therefore, they are unpredictable. This makes the learned behaviour more resistant to extinction, ie. stopping it, because the person continues responding, or “rolling the dice”, expecting that the next time might bring the reward. The brain’s reward circuits that involve dopamine become increasingly sensitised because the unpredictability heightens anticipation and motivation causing dopamine spikes to occur[i]. This also explains why social media can be so addictive. The unpredictability of whether a post or selfie will get liked, shared, or commented on can drive compulsive checking.
In a similar fashion, when people search for information online, it also activates the reward system because finding information feels rewarding. Our brains are wired through evolution to seek out useful information as a survival skill, and the Internet amplifies this drive because searching and clicking for more information stimulates a dopamine-driven feedback loop. The positive, partial reinforcement of searching behaviour is similar to gambling as the seeking system is a dopamine-energised, mesolimbic system[ii]. This also explains addictive scrolling on social media sites. For instance, the pull-to-refresh mechanism on smartphones and tablets which may or may not reveal a new notification on release[iii].
Many ecommerce interfaces have incorporated these psychological hooks into their design. Through partial reinforcement, our brains release dopamine each time we “win” which could be buying a limited or popular item, finding a flash sale, or getting the best price for something. This unpredictable reward keeps us scrolling, clicking, and buying. Which brings me to digital nudges.
Nudges in marketing were derived from Nudge Theory made popular in the Richard H. Thaler and Cass R Sunstein’s 2008 book, Nudge: Improving decisions about health, wealth and happiness. The initial concept came from behavioural economics which explains how subtle changes in the way choices are presented can influence a person’s decisions and behaviours. In essence, a nudge works by designing the choice environment in a way that encourages better or more desirable outcomes for the individual or for society without limiting freedom of choice. The psychology behind nudges is that humans don’t make rational decisions but rely on mental shortcuts or heuristics that drive “rule of thumb” decisions. In ecommerce, digital nudges are built into the user experience (UX) and conversion funnels to drive desired commerce actions, such as adding to a cart, completing a checkout, or subscribing. The most common nudges used are incorporating default settings, or including prompts based on psychological heuristics such as social proof, loss aversion, scarcity effect, decoy effect and anchoring within the customer journey [iv].
Digital nudging walks a fine line between persuasion and manipulation making ethical considerations important when designing automated triggers and workflows to guide online user choice. This is even more important when using agentic AI.
Interactions between people and intelligent software agents (ISAs) such as recommender systems or digital assistants are ubiquitous. ISAs often act as mediators between people and the content or products they seek. These agents are far from neutral and have been modelled with specific goals and reward systems that depend on how users behave. This forms a feedback loop where the ISA learns and acts to influence user choices in ways that maximises its own objectives. In effect its own version of reinforcement, forming its own heuristics, which may or may not match the user’s best interests. This results in the agent either competing or collaborating with the user. To induce the user to perform certain actions the ISA may use persuasion, deception or coercion. The ISA is then rewarded by outcomes such as click-through-rates, downloads, views[v].
According to (Burr et al., 2018) examples of coercion are:
- When a user seeks information but cannot skip a promotional video
- When a user needs to share certain personal data to access content
Examples of deception are:
- Clickbait
- Unwanted pop-ups
Persuasion, which is when a user and ISA enter into a more collaborative interaction, takes the form of either nudging, trading or second order effects. Nudging techniques have been described earlier in the article. Trading takes place when an ISA receives a signal from a user’s online action and offers choices that are expected to benefit both parties. In these exchanges the ISA uses a model of the user’s preferences, for example behavioural segmentation or visiting a certain webpage, to guide its actions. This is known as collaborative filtering. But the ultimate aim remains to maximise its own utility, and the user’s satisfaction serves as a means to achieve that goal. Trading may take the form of financial incentives such as discounts or dynamic pricing. Second order effects refer to unintended long-term effects of interaction with ISAs. Instead of just shaping short-term decisions or behaviours, these interactions can gradually lead to lasting psychological or behavioural shifts. Over time interactions may create feedback loops that influence a user’s beliefs, preferences or sense of value. For example, repeated exposure to specific types of content can reshape what users believe to be normal or widely accepted views, while constant rewarding stimuli can foster addiction, altering what a user finds valuable or desirable.
Where does this leave marketers who aim to increase revenue through ecommerce sites? Here are some guidelines to ensure digital nudges respect user autonomy while promoting transparency, fairness and mutual benefit between platform and user.
- Preserve user autonomy
Design nudges that empower rather than manipulate. Users should always have the freedom to choose, with clear options to decline or adjust default settings. Avoid deceptive “dark patterns” that pressure users into decisions they wouldn’t otherwise make. - Promote transparency and informed consent
Clearly communicate when AI agents personalise recommendations or prices. Users should understand how and why certain options are being presented, and what data informs those suggestions. - Align business goals with consumer well-being
Ensure that nudging strategies create shared value. Encourage choices that are beneficial, sustainable, or genuinely aligned with user interests rather than optimising solely for engagement or sales. - Ensure accountability in AI-driven nudging
Intelligent agents should be designed with ethical guardrails that prevent them from reinforcing harmful or addictive behaviour patterns. Regular audits of AI logic and reward systems can help ensure alignment with ethical standards. - Use behavioural insights responsibly
Apply cognitive and behavioural principles such as social proof, anchoring, or scarcity with honesty and restraint. The goal is to clarify decision-making, not to exploit psychological biases. - Provide feedback and opt-out mechanisms
Give users control over their experience. Allow them to adjust recommendation settings, disable nudging features, or report when interactions feel intrusive or manipulative. - Embed ethical design and oversight
Integrate ethical review into UX design and AI development processes, drawing on interdisciplinary expertise from behavioural science, psychology, and ethics. Frameworks such as Privacy by Design and Responsible AI can provide structure and accountability. - Monitor long-term effects
Evaluate not just short-term conversion metrics but also long-term impacts on user trust, satisfaction, and well-being. Watch for unintended second-order effects, such as dependency or shifts in user beliefs driven by algorithmic reinforcement.
Digital nudging is a useful and effective tool for commercial outcomes, however, when modelled on agentic AI, a balance between persuasive algorithms and ethical design is necessary. When designed with transparency, respect and responsibility digital nudges can enhance both the consumer experience and commercial success while fostering trust rather than manipulation.
[i] Coon, D., & Mitterer, J. O. (2012). Introduction to psychology: Active learning through modules (12th ed.). Wadsworth/Cengage Learning.
[ii] Parsons, T. D. (2017). Cyberpsychology and the brain : The interaction of neuroscience and affective computing. Cambridge University Press. https://doi.org/https://doi.org/10.1017/9781316151204
[iii] Burr, C., Cristianini, N., & Ladyman, J. (2018). An analysis of the interaction between intelligent software agents and human users. Minds and Machines : Journal for Artificial Intelligence, Philosophy and Cognitive Science, 28(4), 735-774. https://doi.org/10.1007/s11023-018-9479-0
[iv] Schneider, C., Weinmann, M., & Brocke, J. v. (2018). Digital nudging: Guiding online user choices through interface design. Communications of the ACM, 61(7), 67-73. https://doi.org/https://doi.org/10.1145/3213765
