How to Win at Angry Birds
← Last Lesson | Next Lesson →
Lesson 03 - The Comparison
Transcript
So let's explore this distinction between game-level interventions and source code interventions a bit further, particularly as it pertains to therapeutics, in order to further understand the difference between evolutionarily familiar and evolutionarily novel health interventions. Another term for game-level interventions would be naturalistic or evolutionarily familiar with respect to the nature of the intervention. Along with how and where it impacts a biological system, in this case, the human body. By definition, all of the evolutionary forces that have shaped our biology and Physiology are at the game level.
In other words, these forces represent the environmental inputs to which the human body has been responding and adapting to throughout the history of our species. So you can also think of game-level actions as encompassing all of the possible moves that we can make in the game of life. Source code interventions, on the other hand, are evolutionarily novel, representing forces that have not shaped our biology and Physiology and thus to which we have no specific adaptations to. This distinction leads to some very important differences between these two kinds of interventions, first, since they are evolutionarily familiar.
Game level interventions are inherently safer and much less likely to crash the system because they engage the regulatory mechanisms that have evolved over millions of years in response to them. Source code interventions, being evolutionarily novel, are also inherently riskier because they bypass evolved regulatory mechanisms. On some level, we recognize this to be true. We require extensive safety testing of synthetic pharmaceuticals for this very reason. But we don't have similar safety concerns about game level interventions. Like exercise, game level interventions are also inherently more powerful because their locus of action is upstream in the physiological cascade and therefore influence the activity of many pathways.
On the other hand, source code interventions are inherently weaker because their locus of action relative to game level interventions is downstream and their scope of influence is narrower. Game level interventions are thus better suited for intervening in complex adaptive systems like the human body, where the range of effects of source code interventions cannot be predicted. In fact, even with complete knowledge and understanding of human biology, which we are still very far from, we'd still be unable to predict the outcome of source code interventions similar to fixing a typo in the games code. Source code interventions are most appropriate for conditions that are clearly caused by a single identifiable factor, such as the meningococcus bacteria.
That's a cause of bacterial meningitis, and it is no coincidence that essentially all major progress in pharmacological treatment has been in the treatment of acute binary single factor conditions. So to review the differences between game level interventions and source code interventions, game level interventions are evolutionarily familiar. While source code interventions are evolutionarily novel, game level interventions are inherently safer because they engage evolved regulatory mechanisms. Source code interventions are inherently riskier because they bypass evolved regulatory mechanisms.
Game level interventions locus of action is upstream, which gives them a much greater scope of influence. The locus of action for source code interventions is downstream. Giving them a narrower scope of influence. Game level interventions are best suited for intervening in complex adaptive systems, whereas source code interventions are better suited for intervening in complicated systems where the effects of that intervention are predictable. This distinction also matters in terms of how we advance therapeutic knowledge, or the proper way we would go about figuring out whether a game level intervention or a source code intervention works or not.
The present gold standard for medical interventions is the randomized controlled trial. The problem here is that randomized controlled trials are designed to evaluate static, discontinuous interventions that impact a single variable. In other words, randomized controlled trials are an appropriate way to determine the effectiveness of source code interventions. Game level interventions, however, are dynamic and continuous by nature, impacting multiple variables simultaneously. The randomized controlled trial is thus not the right way to evaluate them. As mentioned, they are designed to evaluate interventions of a very different sort.
To help in understanding why randomized controlled trials are inappropriate here, imagine a baseball player forced to learn how to throw a curveball by adjusting 1 variable at a time. Not only would this approach to learning to throw a curveball be extraordinarily time consuming and tedious. It would virtually assure that he or she would never arrive at the optimal solution as it can't assess the contextual dependencies of all the variables. And this matters because one reason why game level interventions are under researched and underutilized in healthcare today is because we do not have a research ecosystem for advancing our understanding of how to best use them.
Furthermore, we created a criteria for validating their use, the randomized controlled trial that is impossible for them to meet. It's often said that the reason diet and lifestyle interventions aren't more widely used is because we need more trials, specifically more randomized control trials, since this has become the de facto standard for validating in new therapy. But the very reason we don't have more randomized control trials for them is because they are the wrong tool for the job.
We've created an impossible standard, the only way to advance game level interventions in modern medicine according to existing criteria. Is to conduct randomized controlled trials on them. Yet by design, randomized controlled trials can't advance our knowledge of game level interventions. So how do we learn to play the game, in this case the game of help? Game level knowledge in any complex adaptive system is acquired by trial and error by playing the game as it's designed. That means we make a move, assess the outcome of that move based on the feedback we receive, modify our approach, and make another move.
The truth is, there are many kinds of knowledge we can acquire that don't require randomized controlled trials. In fact, almost all of the knowledge that you currently possess, including things that you consider unassailable truths about the world, were acquired through this virtuous feedback cycle. For example, we all accept that walking off a Cliff is bad for our health, and yet we've never done a randomized trial to prove that this is true.
And if someone were to argue that, we couldn't make this claim without a randomized trial. Or that suggesting that walking off a Cliff was bad for your health wasn't based on nothing but flimsy anecdotal evidence. We wouldn't take them seriously. We know this should be true for other reasons, from our knowledge of basic principles like gravity and the forces that a human body can withstand, and our prior personal experience and so on.
The point being, there are many forms of knowledge that we can construct with confidence without the need for a randomized trial. Another important implication is that understanding the mechanisms of how an intervention works is not required to utilize that intervention. In fact, not only do you not need to know why a game level intervention works at a mechanistic level in order to use it effectively, but knowing how it works rarely improves gameplay. For example, a pitcher can throw a curveball with no explicit knowledge of the physics involved. And furthermore. Having explicit knowledge of the physics involved doesn't really help to throw a better curveball.
Last I checked, there were no physics professors in Major League Baseball or back to our parable. You needn't know anything about how the game works to win the Angry Birds Championship, as team source code learned, and yet our prevailing approach to finding health interventions over the past half century. Has been to start by identifying the mechanisms of disease based on the idea that this is how we will ultimately win the game. So we have different forms of knowledge, game level and source code, both of which are valuable, both which have their own particular domain of application of where they are most useful. .
But presently almost all of our research dollars and efforts are being spent towards acquiring the wrong kind of knowledge for how to treat chronic disease.
Outline
Game-Level vs. Source Code Interventions
Definitions and Distinctions
- Game-Level Interventions: Evolutionarily familiar actions; safe due to engagement with evolved regulatory mechanisms.
- Source Code Interventions: Evolutionarily novel; riskier due to bypassing evolved mechanisms.
Characteristics
- Safety and Power:
- Game-Level: Safer, more powerful due to upstream action affecting multiple pathways.
- Source Code: Riskier, weaker due to downstream action with narrower influence.
Suitability
- Complex Systems: Game-Level interventions ideal for complex systems like the human body.
- Single Factor Conditions: Source Code interventions suitable for acute, binary conditions (e.g., bacterial meningitis).
Research and Evaluation Challenges
Randomized Controlled Trials (RCTs)
- Source Code: RCTs suitable for evaluating static, single-variable interventions.
- Game-Level: Dynamic, multi-variable; RCTs not an appropriate evaluation tool.
Learning and Knowledge Acquisition
- Trial and Error: Essential for understanding complex systems and game-level interventions.
- Mechanistic Understanding: Not necessary for effective use of game-level interventions.
Implications for Therapeutic Advancement
Current Research Focus
- Predominantly on source code knowledge, despite its limited applicability to chronic diseases.
Need for Paradigm Shift
- Emphasis on game-level knowledge through experiential learning and trial and error, bypassing the limitations of RCTs for complex interventions.
Summary
- Game-Level interventions align with evolutionary biology, offering safer and broader impact due to their compatibility with human physiology.
- Source Code interventions, while crucial for specific acute conditions, pose higher risks and have limited scope due to their novel nature.
- The prevailing medical research model, focused on RCTs and mechanistic understanding, is misaligned with the nature of complex adaptive systems, necessitating a shift towards a more holistic, experiential approach to health and therapeutics.
Quiz Yourself
- What distinguishes game-level interventions from source code interventions in the context of health therapeutics?
- Game-level interventions are based on modern technology, while source code interventions use traditional methods.
- Game-level interventions are evolutionarily familiar and safer, while source code interventions are novel and riskier.
- Source code interventions can be applied to any condition, while game-level interventions are only for physical fitness.
- Game-level interventions require extensive safety testing, whereas source code interventions do not.
- Why are randomized controlled trials (RCTs) not suitable for evaluating game-level interventions?
- Because RCTs are too expensive and time-consuming to carry out.
- Game-level interventions affect multiple variables simultaneously, making RCTs designed for single-variable interventions inappropriate.
- All game-level interventions have already been proven effective without needing RCTs.
- RCTs are only suitable for pharmaceutical interventions, not lifestyle or dietary changes.
- How can knowledge about effective game-level interventions be acquired according to the text?
- Exclusively through conducting large-scale epidemiological studies.
- By trial and error, assessing outcomes, and adjusting approaches based on feedback.
- Only through understanding the mechanistic details of how interventions work.
- Through anecdotal evidence and personal testimonials only.
- B
B
B