Algorithms to Live By — Interactive Mindmaps

Key concepts: 1. Optimal Stopping: When to Stop Looking

1. Optimal Stopping: When to Stop Looking

The Secretary Problem Framework

• Universal dilemma of when to stop searching and commit
• Goal: hire the single best candidate from a sequential, random-order interview process
• Key constraints: only relative comparisons, no recall of passed candidates, immediate decision ends search
• Origins are murky, popularized by Martin Gardner in 1960 with roots in 19th-century suitor puzzles

The Optimal Strategy: Look-Then-Leap Rule

• Two-phase approach: initial 'look' phase to gather information, then 'leap' phase to commit
• Mathematically optimal to switch after reviewing 37% of candidates (the 37% Solution)
• Yields a constant 37% chance of success, regardless of pool size
• Strategy balances exploration (setting a benchmark) and exploitation (acting before best is lost)

Real-World Applications & Complications

• Applied to romance: calculating when to start seriously committing to a partner
• Real life introduces rejection (candidate says no) and recall (returning to a past option)
• With 50% rejection chance, optimal look phase shrinks to 25% of search
• With recall allowed, look phase extends to 61% with fallback to best past option

Full Information & The Threshold Rule

• When objective data is available (e.g., percentile scores), strategy shifts
• Use a sliding percentile threshold: hire anyone above a predetermined score
• More efficient than the Look-Then-Leap Rule under full information
• Applies to scenarios where candidates can be quantitatively ranked

Optimal Stopping in Life Decisions

• Selling a house: maximize profit by weighing each offer against cost of waiting
• Set a single price threshold in advance and never revisit rejected offers
• Parking spot search: occupancy rate dictates when to start seriously looking
• Cities can optimize parking via pricing based on optimal stopping principles

The Decision to Quit & Problematic Scenarios

• Burglar problem: simple formula for when to stop (attempts = success chance / failure chance)
• Some scenarios (e.g., 'triple or nothing' gamble) have no safe stopping point and should be avoided
• Highlights that not all sequential decision problems have a mathematically sound optimal stop

Human Behavior vs. Mathematical Optima

• Research shows humans consistently stop searching earlier than math advises
• This may be rational due to endogenous costs: time, effort, boredom, and opportunity cost
• Pure mathematical models ignore psychological and real-world search costs
• Optimal stopping theory provides a framework, not a strict prescription, for irreversible choices

Full Information and the Threshold Rule

• Objective data like percentile scores eliminates the need for a calibration phase in decision-making
• A Threshold Rule is used: hire immediately if an applicant exceeds a specific percentile threshold
• The threshold depends on how many candidates remain—higher standards with more options, lower with fewer
• This approach boosts success rates to 58%, showing clear metrics make decisions more efficient

Selling with Costs: The House Selling Problem

• Unlike the secretary problem, selling involves known dollar values and real costs like mortgage payments
• Goal is to maximize total profit by balancing better future offers against immediate waiting costs
• Optimal stopping price depends only on cost per offer and known offer range—set threshold in advance
• Never recall a past offer—if it was below threshold initially, it remains below due to sunk costs

Parking as Optimal Stopping

• Core variable is occupancy rate—percentage of spots filled—which creates cruising when too high
• Look-Then-Leap Rule applies: pass all spots beyond critical distance, then take first one after
• Critical distance shrinks dramatically as occupancy decreases (quarter-mile at 99% vs. half-block at 85%)
• Adaptive pricing to target ~85% occupancy simplifies the stopping problem and reduces societal costs

Quitting and the Burglar Problem

• Mathematical framing of 'quit while you're ahead' using sequential risk-reward decisions
• Optimal number of attempts is roughly chance of success divided by chance of failure
• Some scenarios like 'triple or nothing' have no optimal stopping rule—mathematically advise playing forever
• Highlights that some risk scenarios are best avoided entirely rather than optimized

Human Psychology and Optimal Stopping

• People consistently stop too early in laboratory secretary problems, leaping sooner than optimal
• This impatience may be rational when considering endogenous time costs like boredom and mental toll
• Early stopping aligns with rationally adjusted models that account for human factors
• Optimal stopping mirrors the nature of time itself—one-way sequence of irreversible decisions

The House Selling Problem (Known Distribution)

• Calculate a single, rational threshold upfront based on known offer distribution and constant search costs.
• Hold this threshold firmly; do not lower it out of impatience as time passes.
• Never revisit a rejected offer; the optimal strategy is forward-looking only.

The Parking Problem (Spatial Search)

• Strategy is determined by the occupancy rate (density of available options).
• Higher vacancy rates allow for greater choosiness and searching closer to the destination.
• The optimal strategy balances the risk of passing a good spot against the cost of circling back from farther away.

Sequential Risk and the Quit Formula

• Applies to scenarios with repeated attempts and a chance of catastrophic failure (total loss).
• A simple heuristic: quit after (probability of success) / (probability of failure) attempts.
• This rule balances the diminishing chance of eventual success against the accumulating risk of ruin.

Unwinnable Games (e.g., Triple or Nothing)

• Some scenarios have no mathematically optimal stopping point for long-term gain.
• Indefinite play is mathematically destined for ruin due to the risk structure.
• The rational strategy is complete avoidance, not attempting to find a stopping rule.

Human Behavior vs. Mathematical Optima

• People tend to stop searching earlier than pure mathematical models prescribe.
• This is often due to real but unquantified costs like time, effort, and psychological fatigue (boredom).
• Such early stopping is not necessarily irrational but highlights the need to balance formal models with experiential costs.

Key concepts: 2. Explore/Exploit: The Latest vs. the Greatest

2. Explore/Exploit: The Latest vs. the Greatest

The Core Explore/Exploit Trade-off

• Universal tension between trying something new (explore) and sticking with a known favorite (exploit)
• Exploitation enables rich, rewarding experiences (e.g., traditions, beloved books)
• Pure exploration can be a curse, preventing enjoyment of known good options
• Life presents constant 'slot machines' with unknown odds requiring strategic balance

The Multi-Armed Bandit Problem

• Classic computer science formulation of the explore/exploit dilemma
• Named for slot machines ('one-armed bandits') with unknown payout odds
• Goal: maximize total winnings through optimal mix of trying and playing machines
• Challenge: assessing limited data with high uncertainty about true value

Time Horizon as Critical Factor

• Optimal strategy depends entirely on remaining time interval
• Long future favors exploration to find better options
• Short remaining time favors exploitation of known best options
• Value of exploration decreases as time to enjoy discoveries runs out
• Observed strategies reveal perceived intervals (e.g., Hollywood's sequel-heavy approach)

Mathematical Solutions & Algorithms

• Early approach: Win-Stay, Lose-Shift (simple but flawed, too rash after single failure)
• Breakthrough: Gittins Index assigns single score blending known value and unexplored potential
• Gittins mathematically justifies 'exploration bonus' for unknown options
• Practical focus: minimizing regret (pain of missed opportunities)
• Upper Confidence Bound (UCB) algorithms: choose option with highest plausible upside

Real-World Applications & Ethical Dimensions

• A/B testing: treats users as slot machines to optimize clicks and sales
• Clinical trials: ethical dilemma between gathering knowledge for future patients vs. best care for current participants
• Human tendency to over-explore beyond mathematical optimum
• 'Restless world' justification: continued exploration rational as options change over time

Life Arc Through Explore/Exploit Lens

• Childhood as protected exploration phase: play and curiosity gather world information
• Aging brings strategic shift toward exploitation as perceived time becomes limited
• Social circle pruning: focusing on deep, rewarding relationships
• Exploitation in later life can lead to greater satisfaction

The Gittins Index Breakthrough

• John Gittins solved the bandit problem by introducing a 'bribe' or guaranteed payout rate that would make one willing to abandon a slot machine forever.
• The optimal strategy is to always play the arm with the highest current Gittins Index, which elegantly balances exploration and exploitation in a single metric.
• The index assigns a substantial 'exploration bonus' to unknown options, mathematically justifying why 'the grass is always greener'—the potential of the unknown has quantifiable future value.
• It confirms 'win-stay' logic and challenges 'lose-shift' reasoning, showing that losing on a well-understood machine doesn't always mean you should abandon it.

Limitations of Theoretical Solutions

• The Gittins Index relies on strict assumptions like geometric discounting, which doesn't always align with human psychology or real-world decision-making.
• It becomes computationally impractical in many real-life scenarios and falls apart if there's any cost to switching between options.
• Perfect mathematical solutions are often too rigid for fluid, everyday explore/exploit dilemmas, prompting the search for simpler, more flexible strategies.

Regret as a Guiding Metric

• Regret is defined as the difference between the payoff received and the payoff that could have been achieved with perfect hindsight.
• Even with the best strategies, total regret will always increase over time when information is imperfect, but a good strategy makes it increase at a slowing rate.
• The theoretical minimum is for regret to increase logarithmically, meaning you make as many mistakes in your first ten tries as in the next ninety—a mathematical consolation that new regrets diminish over time.

Upper Confidence Bound (UCB) Algorithms

• UCB algorithms operate on the principle of 'optimism in the face of uncertainty,' choosing the option with the highest plausible upside rather than the highest known average.
• They calculate a confidence interval for each option and act as if the true value is at the top of that range, naturally boosting less-tested options.
• This approach automatically balances exploration and exploitation, providing a formal justification for giving new experiences the benefit of the doubt.

A/B Testing and the Digital Bandit

• The explore/exploit problem powers A/B testing on the modern internet, where companies like Google and Amazon run continuous experiments to optimize user engagement and revenue.
• In this model, users are not the gamblers but the slot machines being studied, with their attention and wallets as the 'jackpot.'
• A canonical success story is Barack Obama's 2008 campaign, where A/B testing the donation page raised an additional $57 million by tailoring messages to different audience segments.

Ethical Tensions in Medical Clinical Trials

• Standard randomized trials knowingly assign some patients to inferior treatments to definitively determine efficacy.
• Adaptive trials, modeled on bandit algorithms, adjust treatment probabilities in real time to give more patients the better treatment during the experiment.
• The ECMO trial controversy highlighted the tragic conflict between gaining knowledge for future patients and providing best care to current trial participants.
• The multi-armed bandit framework forces difficult conversations about minimizing regret in human lives, not just clicks and revenue.

The Shift Toward Adaptive Clinical Trials

• Statistician Don Berry applied bandit principles to design adaptive cancer trials at MD Anderson.
• His critique of rigid randomized trials is gaining traction in medical research.
• The FDA's recent draft guidance endorsing 'Adaptive Design' trials signals potential regulatory shift.

Human Tendency to Over-Explore

• Laboratory studies show people consistently explore more than mathematically optimal.
• In classic experiments, participants spent far more time observing than optimal before switching to exploitation.
• Only 30% of people use near-optimal strategies in bandit tasks, while most use simpler, exploration-heavy approaches.
• People fail to make clean breaks from underperforming options and continue unnecessary alternation.

The Challenge of a Restless World

• Real-world 'restless bandit' problems have no tractable optimal solution as payoffs change over time.
• Continued exploration becomes rational necessity when restaurant quality or airline reliability can fluctuate.
• Algorithms like Gittins index offer approximations in stable environments.
• Evolutionary instincts for fluctuating worlds may be mismatched in era of standardized products.

Childhood as Evolutionary Exploration Phase

• Childhood is an evolutionary solution to the explore/exploit trade-off.
• Extended dependence allows children to explore wildly without pressure to exploit for payoff.
• What appears as cognitive deficiency (poor planning, lack of focus) is actually optimized information-gathering.
• Caregivers handle exploitation needs, freeing children for risk-free learning.

Old Age as Strategic Exploitation Phase

• Elderly deliberately prune peripheral relationships to focus on meaningful core connections.
• This strategic choice is driven by heightened awareness of limited time.
• Experiments show this shift relates to perceived 'interval' rather than age itself.
• The shift from exploration to exploitation helps explain increased life satisfaction with age.

The Life Arc Through Explore/Exploit Lens

• The framework reveals rational structure behind behaviors from toddler curiosity to elder routines.
• Human development follows strategic pattern: childhood exploration followed by adult exploitation.
• Perceived time horizon drives strategic shifts in social behavior and decision-making.
• The dilemma provides powerful lens for understanding entire human lifespan.

Key concepts: 3. Sorting: Making Order

3. Sorting: Making Order

The Fundamental Nature of Sorting

• Sorting is a core, ubiquitous, and computationally difficult task that spurred the invention of early computing
• It defines our experience of information as the universal user interface for search engines and digital lists
• Suffers from diseconomies of scale - becomes disproportionately harder as the problem grows larger

Algorithmic Strategies and Efficiency

• Big-O notation categorizes algorithms by how their effort grows with problem size
• Intuitive methods like Bubble Sort and Insertion Sort fall into inefficient quadratic time
• Divide and conquer algorithms like Mergesort achieve faster linearithmic time
• Bucket Sort uses categorization to achieve lightning-fast linear time by exploiting known properties

The Sorting-Searching Trade-off

• Sorting is only worthwhile if you plan to search the data frequently
• For personal tasks, searching digitally is often cheaper than manual sorting
• Large-scale operations like Google justify massive upfront sorting costs

Sorting in Social and Competitive Systems

• Sports leagues function as algorithms for ranking teams through different tournament structures
• Round-robin tournaments act as thorough but slow Comparison Counting Sorts
• Single-elimination brackets are fast but brittle in noisy competitive environments
• Season-long sorting provides more robust rankings than knockout playoffs

Organic Sorting in Biological and Social Systems

• Dominance hierarchies in animal groups form through decentralized, pairwise interactions
• Maintaining pecking orders carries significant cognitive and physical burdens that scale poorly
• Humans achieve efficient sorting through shared consensus and cardinal measures
• Shifting from ordinal to cardinal sorting allows civilization to organize peacefully at massive scale

Big-O Notation: Classifying Algorithm Efficiency

• Describes how an algorithm's run time grows with input size (n), focusing on worst-case scenarios
• Classifies algorithms into families: O(1) constant time, O(n) linear time, O(n²) quadratic time, and O(n!) factorial time
• Intentionally ignores constant factors to focus on dominant scaling relationships as n becomes large

Quadratic Sorts: Intuitive but Inefficient Methods

• Bubble Sort repeatedly scans and swaps adjacent items, requiring n passes in worst-case scenarios
• Insertion Sort builds sorted lists by inserting items into correct positions among already-sorted elements
• Both methods become cripplingly slow for large datasets due to n² scaling, making them impractical for large-scale sorting

Divide and Conquer: The Mergesort Breakthrough

• Uses recursive splitting of lists into halves until reaching single items (which are sorted by definition)
• Achieves O(n log n) linearithmic time by merging already-sorted lists in linear time operations
• Represents monumental improvement over quadratic sorts, especially for large datasets like census-sized collections

Bucket Sort: Linear-Time Sorting Without Comparisons

• Avoids the O(n log n) comparison-based limit by categorizing items into ordered buckets based on known properties
• Enables linear-time sorting as demonstrated by library systems like the Preston Sort Center processing 85,000 books daily
• Requires knowledge of data distribution to create appropriately sized buckets for optimal efficiency

Hybrid Human-Machine Sorting Strategies

• Expert human sorters like librarians use mental bucket sorting based on experience with data patterns
• Combines Bucket Sort for creating initial order with Insertion Sort for final refinement of small groups
• Represents gold standard approach ratified by both human expertise and machine optimization principles

The Sorting vs. Searching Trade-Off

• Sorting is only valuable as preemptive investment against future search operations
• Justified when search time costs exceed sort time costs and repeated searches are anticipated
• Digital search tools have reduced value of manual organization, making 'messy' approaches sometimes optimal

Sports Tournaments as Sorting Algorithms

• Round-robin tournaments function as comparison counting sorts with O(n²) game requirements
• Ladder tournaments operate like Bubble Sort with slow, incremental sorting processes
• Bracket tournaments (like March Madness) use Mergesort-like structures but single elimination only finds champions, leaving other teams unsorted

The Economics of Sports Scheduling

• Sports leagues prioritize maximizing suspense and revenue over minimizing games.
• Season structures are deliberately designed to delay resolution and maintain fan interest.
• The games themselves are the primary product, not just a means to determine a champion.

Noise and the Fragility of Champions

• Real-world comparisons are noisy—a better team doesn't always win.
• In noisy single-elimination tournaments, even dominant teams have low championship probabilities.
• Fast sorting algorithms like Mergesort are brittle to erroneous comparisons, while slower ones like Bubble Sort are more robust.
• Comparison Counting Sort (analogous to a Round-Robin season) is the most robust against noise.
• A team's regular-season performance is a truer measure than playoff elimination, which often involves luck.

Organic Sorting in Social Systems

• Sorting can emerge organically from individual interactions, as seen in poker or animal hierarchies.
• Dominance hierarchies in animals (e.g., macaques) use displacement to maintain stable social order.
• These systems are bottom-up sorting algorithms that minimize violent conflict over resources.
• Online poker tables exhibit similar jockeying, with lower-ranked players displaced by better ones.

The Computational Burden of Pecking Orders

• Pecking orders emerge as violent but efficient solutions to establishing hierarchy.
• Confrontations serve as comparisons that resolve rankings and preempt further violence.
• Disrupting natural sorting procedures (e.g., debeaking chickens) can increase overall antagonism.
• Aggressive acts per individual rise with group size but diminish once hierarchy is settled.
• Dominance hierarchies are information hierarchies requiring detailed mental maps to minimize conflict.

From Confrontations to Consensus

• Decentralized sorting imposes a cognitive load of tracking shifting rankings.
• Humans achieve more efficient sorting through shared consensus (e.g., poker player rankings).
• Internal ranking agreement reduces the need for constant competition.
• Conflicts arise when internal rankings diverge, highlighting the role of communication and memory.

The Efficiency of Races Over Fights

• Sorting via races (e.g., marathons) is a constant-time process compared to pairwise matchups.
• Cardinal ranking (using measurable performance) transforms the computational burden versus ordinal ranking (direct comparisons).
• Competing against a standard reduces physical risk and mental strain as group size scales.
• Sports like skiing or running allow athletes to be sorted without direct confrontation.

Cardinal Measures in Social Hierarchies

• Cardinal measures (money, size, age) simplify status determination in complex societies.
• Examples include the Fortune 500 (revenue) and maritime 'Law of Gross Tonnage'.
• In nature, size-based dominance (e.g., in fish) avoids the bloodshed seen in chickens or primates.
• Metrics like GDP or respect for elders convert potential conflicts into straightforward rankings.
• The shift from ordinal to cardinal sorting is a cultural innovation enabling large-scale cooperation.

Pecking Orders as Decentralized Sorting

• Animal hierarchies are established through pairwise confrontations, functioning like a decentralized sorting algorithm.
• The initial violence serves to establish a stable order that preempts future conflict over status.
• Disrupting the established order, such as by removing a top individual, can trigger renewed aggression as the hierarchy is recalculated.

Computational Cost of Social Sorting

• The mental effort required to sort individuals grows with group size, as each must track their rank relative to others.
• Humans mitigate this cost through consensus and shared cultural knowledge, which provides a pre-agreed ranking system.
• Efficient sorting in large groups requires moving beyond constant pairwise comparisons to more scalable methods.

Races and Cardinal Measures for Efficient Sorting

• Races transform sorting from O(n²) pairwise comparisons into O(n) operations by using a shared benchmark like time or distance.
• Cardinal measures such as money, test scores, or job titles provide constant-time sorting metrics.
• These quantifiable benchmarks enable peaceful, scalable order by replacing physical confrontations with objective comparison.

Human Societal Benchmarks and Order

• Human societies use constructed benchmarks (e.g., wealth, rules, laws) to simplify complex status hierarchies.
• These systems dramatically reduce conflict by providing clear, often non-violent, criteria for sorting and status.
• The use of shared benchmarks is a key distinction enabling the large-scale, cooperative organization of human groups.

Key concepts: 4. Caching: Forget About It

4. Caching: Forget About It

The Universal Dilemma of Limited Space

• Managing limited storage presents the same fundamental questions in both physical and digital contexts
• The core challenge is deciding what to keep and how to arrange it efficiently
• Traditional organization methods (like grouping similar items) may not be optimal compared to computer science principles

The Memory Hierarchy

• Computers use a pyramid structure where each level is larger but slower than the one above it
• This solves the trade-off between storage capacity and access speed
• First practical implementation was in the 1962 Atlas supercomputer with fast working memory
• Analogous to keeping frequently used books on a desk rather than in a distant library

The Cache Concept

• A cache is a fast storage pool that proactively retains data likely to be needed again
• Invented by Maurice Wilkes to prevent slow trips back to main storage
• Critical for overcoming the 'memory wall' between processor speed and memory access
• Now exists at every level of computing from processors to web browsers

Cache Eviction Policies

• When cache is full, an eviction policy determines what to remove
• Theoretically optimal policy is Bélady's Algorithm (clairvoyant future knowledge)
• Since clairvoyance is impractical, systems must approximate optimal behavior
• Most effective practical approach is Least Recently Used (LRU)

Least Recently Used (LRU) Principle

• Evicts the item untouched for the longest time
• Leverages temporal locality (recently used items tend to be needed again soon)
• Applied beyond computers in application switchers and library systems
• Can revolutionize physical spaces like libraries by creating prominent displays of recently returned books

Geographic Caching and CDNs

• Addresses distance problems, not just speed problems
• Content Distribution Networks (CDNs) store popular web content closer to users
• Companies like Akamai use this to cut internet delays
• Physical equivalent: Amazon's warehouse organization and anticipatory shipping patents

Caching in Home Organization

• We naturally cache items based on frequency and location of use
• Conscious LRU application means keeping occasionally used items over untouched ones
• Multi-level memory hierarchy optimizes item placement from bedside valet to storage unit
• Arrangement systems like the Noguchi Filing System or simple piles can be mathematically optimal

Self-Organizing Systems and Optimal Arrangement

• Noguchi Filing System: always file and search from the left
• Simple desk piles can be smart, self-organizing structures
• Research proves 'move-to-front' rule (LRU) is incredibly efficient
• Contrary to strict categorization, these systems optimize for retrieval frequency

Human Memory as Caching System

• Hermann Ebbinghaus's forgetting curve shows how memories fade
• Pattern matches statistical decay of word use in language
• Suggests human forgetting is optimal tuning for likely-to-be-needed memories
• The 'tyranny of experience': retrieval slows as mental database grows with age
• Tip-of-the-tongue moments may be natural cost of richer knowledge, not cognitive decline

Philosophical Implications of Caching

• Forgetting is a feature of smart design, not a flaw
• Knowing what to keep close is essential for efficiency
• Principles apply from processors to piles to human minds
• Smart forgetting enables effective remembering across all systems

Several practical policies exist:

• •   Random Eviction: Surprisingly decent, as frequently used items tend to return to the cache quickly.
• •   First-In, First-Out (FIFO): Evicts the oldest item in the cache.
• •   Least Recently Used (LRU): Evicts the item that hasn't been used for the longest time.

Applying LRU to the Library

• Libraries are a natural memory hierarchy.
• They already use LRU-like policies for decisions like which books to send to offsite storage (e.g., those not checked out in over a decade).
• However, they could be far more efficient by fully embracing the cache principle.

Geographic Caching: The Physical Cloud

• Caching isn't only about speed versus cost; it's also about proximity versus distance.
• This is the foundation of the modern internet.
• Companies like Akamai operate global Content Distribution Networks (CDNs), which cache popular website data on servers physically closer to users.

Caching on the Home Front

• The principles of digital caching translate directly to organizing physical spaces.
• Computer architect John Hennessy notes we naturally cache items based on frequency of use, from our desks to deep archives.
• Applying LRU (Least Recently Used) principles means keeping the college T-shirt you occasionally wear, not the plaid pants you haven't touched in years.

Filing and Piling

• Beyond what to keep and where, the final challenge is how to arrange things.
• Contrary to the universal "like with like" advice, economist Yukio Noguchi developed the "Super Organized Method." His system involves filing documents in a single box, always inserting and searching from the left-hand side.
• This simple rule—always returning a file to the leftmost position—means the most recently used items are always the fastest to find.

The Forgetting Curve

• The discussion of memory naturally extends to the human brain.
• Hermann Ebbinghaus's 19th-century experiments mapped the "forgetting curve," showing how memory for nonsense syllables fades over time.
• For over a century, the shape of this curve was a mystery.

Cognitive Decline Reframed: The Cost of Knowledge

• Larger memory stores are inherently slower to search due to increased database size
• Aging brings accumulation of experiences and vocabulary, expanding the mental archive
• Research shows knowing more makes retrieval harder, leading to longer recognition times
• Tip-of-the-tongue moments are often consequences of managing richer knowledge stores
• Occasional cache misses represent a trade-off for lifetime accumulated experience

Optimal Organization Principles

• LRU principle and geographic caching provide intuitive tools for physical organization
• Noguchi Filing System and desk piles are mathematically validated for linear searches
• These methods outperform rigid categorization for frequently accessed items
• Organization should prioritize accessibility of vital information over neatness

Forgetting as Optimal Strategy

• Human forgetting follows patterns tuned to real-world statistical patterns
• Ebbinghaus curve represents an optimal cognitive strategy rather than a flaw
• Forgetting mechanisms help prioritize relevant information over time
• The brain's organization keeps vital information within quick reach despite large stores