"Riku's late again today."
Yuki checked the clock. 4:15 PM. Fifteen minutes past the agreed time.
"As predicted," Aoi said flatly. "Riku's lateness pattern is quite typical."
"Typical?"
"In information theory, there's a concept called typical sequences."
At that moment, Riku burst in, out of breath.
"Sorry! Really sorry today!"
"That's the third time this week," Yuki sighed.
Aoi opened the notebook. "I've been recording Riku's lateness. Data from the past 30 days."
"What, you kept records?" Riku was surprised.
"Look. Late 22 times, on time 8 times. Lateness probability is about 0.73."
Mira quietly approached and looked at the data. She began writing something in her notebook.
"So what's a typical sequence?" Yuki asked.
Aoi started explaining. "In the long term, most sequences converge to a specific pattern. This is the asymptotic equipartition property."
"Sounds difficult..."
"For example, flip a coin 1000 times. Theoretically there are 2^1000 possible results, but actually most results are close to 'about 500 heads, 500 tails.'"
Riku raised his hand. "So my lateness pattern is predictable?"
"In a sense. If late 22 times out of 30, you'll probably continue being late about 73% of the time."
Mira wrote an equation. "-1/n log P(X₁...Xₙ) → H(X)"
"What Mira showed is the definition of typical sequences," Aoi supplemented. "The log-average of a long sequence's probability converges to entropy."
Yuki pondered. "What about unusual patterns?"
"Non-typical sequences. For example, the probability of Riku arriving on time 30 days straight is 0.27^30..."
"Basically zero," Riku grimaced.
"But what's interesting," Aoi continued, "is that the number of typical sequences is only about 2^(nH). n is the sequence length, H is the entropy rate."
"That's few?"
"Yes. While there are 2^n possible sequences, the actually likely typical sequences are exponentially fewer. This forms the basis of source coding."
Mira nodded. She always enjoyed these abstract discussions.
Yuki said, "So reality is more patterned than we think?"
"Yes. Even appearing random, there are statistical regularities. That's the law of large numbers."
Riku's face became serious. "So if I come on time every day from tomorrow, I'd be a non-typical sequence?"
"You would. A probabilistically rare event."
"Alright, I'll try!"
Aoi said calmly, "But once a new pattern is established, it becomes the new typical sequence. The entropy rate just changes."
"Harsh..."
Yuki laughed. "But change is possible, right?"
"Of course. Statistical properties change if you change the generating process. Humans have will."
Aoi drew another diagram. "That's what makes human behavior interesting from an information theory perspective. We're not purely random, not purely deterministic. We exist in between, with patterns that can evolve."
"So we're like adaptive stochastic processes?" Riku asked with surprising insight.
"Exactly that," Aoi smiled.
Mira handed over a small note. "Typical sequences: predictable. Non-typical: surprising. Both exist."
"Mira's right," Aoi said. "Typical daily life and non-typical surprises. Both make information theory and life interesting."
Riku took out his planner. "I'll keep records starting tomorrow. Create my new entropy rate."
"I'm rooting for you," Yuki said.
"But," Aoi added, "human behavior can't be perfectly predicted. That's why mutual information has meaning."
Sunset streamed through the club room window. The information theory club's daily life was typical today too.
But a slightly non-typical hope was mixed in as well.