I put together a few interactive concept notes on bridging AI, NLP, and SimDec. Click the cards below to explore the ideas and their technical breakdowns.
The first idea is something I’ve been thinking about when it comes to the gap between AI predictions and actual business decisions. In areas like marketing, customer behavior is full of uncertainty. Right now, AI models are great at making accurate predictions, but they don’t really help decision makers understand how the various uncertain factors interact.
Since my background is in NLP and customer analytics, I thought it would be really interesting to relate it to your work. The initial idea is to use Transformer models to capture customer behavior, run Monte Carlo simulations, and then apply GSA and SimDec.
What excites me about this isn’t just trying to build another accurate AI. Instead, I’d really like to learn how to use SimDec to break down all that uncertainty created by AI into something clear and actionable that businesses can base their strategies on.
In many marketing decisions like pricing or product positioning, customer behavior is quite uncertain. This becomes even more challenging in areas like sustainable products, where things like environmental awareness or trust in “green” claims can affect decisions in unpredictable ways.
Develop a framework that combines AI-based customer behavior modeling (e.g., using NLP) with Simulation Decomposition (SimDec) to better understand how different uncertain factors interact.
Most AI models are good at prediction but don’t explain how uncertainties work together. Sensitivity analysis shows which factors matter, but not how they interact in shaping decisions.
Moving from mere prediction to a decision-oriented approach, where we explicitly see how different factors jointly affect results to enable better decision-making.
My second thought revolves around pricing in digital markets. This is an area where things get complicated quickly due to so many unknowns. While there are many machine learning models for forecasting demand, most of them focus on just hitting a target with high accuracy. They don’t provide much explanation or tell you what actually drives the risk.
I think it would be interesting to bring SimDec to this area. The approximate approach I have in mind is to set up an ML model to forecast demand, run Monte Carlo simulations, and then apply GSA and SimDec to the results.
Rather than just forecasting a single price point, I’m really excited to see how we can use SimDec to unpack all of that uncertainty. I’d love to work on making these complex pricing models transparent, so that managers can understand the interplay between the various uncertainties before making a decision.
Pricing in digital markets is very sensitive to uncertain factors like demand, customer reactions, and competitor behavior. This makes it hard to make stable and reliable decisions.
Combining machine learning models for demand prediction with SimDec to better understand which uncertainties really drive pricing outcomes and how they interact.
Most pricing models focus on prediction accuracy, but they don’t provide much insight into uncertainty or explainability.
Using SimDec to break down the uncertainty in pricing decisions, making the results more interpretable and genuinely useful for strategic planning.
As my third idea, I was thinking about digital platforms and how heavily they rely on predicting user behavior. The problem is that these predictions inherently carry a lot of uncertainty and, if not managed well, can lead to very poor strategic choices. There’s a lot of focus on making better predictions, but there seems to be much less attention paid to how the uncertainty in these predictions affects the platform’s ultimate decisions.
My idea is to use NLP and behavioral modeling to capture customer behavior, run scenarios through Monte Carlo simulations, and then bring in GSA and SimDec.
What I’d really like to learn and work on here is to shift the focus from building a slightly better predictive model. Instead, I want to explore how SimDec can help us explicitly account for that uncertainty, so that platform managers can see exactly how that uncertainty shapes their strategic decisions.
Digital platforms rely a lot on predicting user behavior, but these predictions are often uncertain, which can lead to poor strategic decisions.
Explicitly model customer behavior accounting for uncertainty, then use SimDec to understand how this uncertainty affects overarching platform-level decisions.
There is a strong focus on prediction accuracy, but less attention on how uncertainty in those predictions practically impacts actual decisions.
Shifting the focus toward decision-making under uncertainty rather than just iteratively improving prediction models.