Background and Motivation

Prediction markets have long been recognized as powerful tools for aggregating dispersed information into probabilistic signals. By allowing participants to express beliefs through market positions, these systems often outperform polls or expert forecasts in estimating the likelihood of discrete events. However, their underlying design assumptions reflect a simplified view of how information and expectations evolve.

Most real-world decision environments are reflexive rather than static. Information does not arrive in isolation, nor do beliefs update instantaneously at the moment an event resolves. Instead, beliefs evolve continuously through interaction between narratives, incentives, social feedback, and algorithmic interpretation. Market prices, media framing, and collective attention often shift before the event they appear to predict, and these shifts can materially influence downstream behavior.

In financial markets, price movements frequently precede fundamental confirmations. In macroeconomics, expectations around policy decisions often reprice assets well before official announcements. In social systems, narratives gain or lose dominance long before their consequences are observable. In AI-mediated environments, model-generated probabilities influence human decisions, which in turn reshape future data distributions.

Despite this, most prediction infrastructures remain outcome-centric. They treat belief formation as an intermediate step rather than a measurable phenomenon in its own right. As a result, existing systems capture what the crowd ultimately believed, but not when, how fast, or through which dynamics that belief emerged.

Preda is motivated by the observation that in many domains, the timing of belief shifts is the primary signal. Understanding when consensus changes—rather than simply whether it does—provides deeper insight into anticipation, coordination, and systemic risk. Time-Shifted Prediction Markets are designed to make this temporal dimension explicit.