questions, problems and experiments Science is often imagined as a logical process—one that begins with an unsolved question or problem, then applies the traditional scientific method, and ends by 248 T HE ENGINE OF SCIENTIFIC DISCOVERY uncovering new experimental findings or a theory. But in reality, we are not able to run an experiment to study what our tools cannot detect—whether using advanced spectrometers or radar telescopes. In most areas of science, if we cannot observe and measure a phenomenon with our tools, it is generally invisible to our imagination too. But some domains offer some exceptions, especially theoretical physics where concepts like gravitational waves and black holes emerged—by using advanced mathematical frameworks like tensor calculus and building on existing discoveries using interferometers and telescopes (Chapter 6). We do science using tools and methods, and those we have at our disposal are what largely define the questions and hypotheses we are even able to ask in the first place. The tools we use determine what phenomena we can explain and predict using them, and what we cannot. But how do tools do this? How do they limit and steer what we can conceive, measure and detect? Take the microscope—it transformed biology not just by discovering cells and bacteria and opening the microscopic world but by enabling us to conceive and tackle entirely new questions. For the first time, we could ask: are cells the basic unit of life? How do they actually function? Do they divide—and if so, how? The answers are essential to the foundation of biology—and these questions would never have arisen without the tool that made cells visible and experiments possible. Inventing x-ray crystallography then revealed the mystery of the structure of DNA by enabling researchers to see its double-helix structure. This unlocked profound new questions in biology: how does DNA's structure replicate itself ? How does it encode and store genetic information? These questions became the foundation of molecular genetics—and they were only possible by inventing x-ray crystallography (Chapter1). Later, creating gene sequencing methods transformed biology again, enabling us to examine genetic variations with unprecedented precision and scale. This new tool sparked completely new types of questions unimaginable before: can we sequence the human genome? How are gene mutations linked to diseases? How can we design treatments tailored to an individual's DNA? Think of new machine learning and AI tools. These have opened new experiments, simulations and ways to detect patterns using datasets so massive that no human could make sense of them alone. From drug discovery to complex systems and climate modelling, AI tools do not just make science faster but let us do an entirely new, deeper kind of science. Innovative AI algorithms help uncover new drug candidates and antibiotics by quickly analysing millions of molecules. They spark fundamental new types of questions: how can we best find hidden patterns in medical records for millions of patients? How can we simulate billions of interactions in an ecosystem or an economy? How can we make more efficient predictions? Take the extraordinary invention of the spectroscope. It did not just give us a new kind of vision but opened transformative new ways to explore the universe. By being able to analyse the light emitted by stars, it opened completely new fascinating questions and answers never imagined before: do the same elements that make up Earth also exist in distant stars? How did these elements form in space and come to be scattered across the universe? Where does the immense heat and energy of stars come T from? What is the chemical makeup of other planets? Surprisingly, we discovered that stars and planets are built from the same fundamental elements on Earth. Different planets could then be seen as a kind of experimental control for how life could emerge. They became natural laboratories, helping us figure out what conditions make life and a liveable planet possible. The universe, at least in part, did not seem so completely distant or alien—it felt more connected, made of the same building blocks. This logic of science—that new tools largely dictate the very scope of scientific inquiry—plays out again and again across all fields. Space telescopes open entirely new astronomical questions about the early universe billions of years ago, about the origins of galaxies and about the behaviour of gravity. Electron microscopes catalyse entirely new research areas in cell biology, materials science and nanotechnology. New tools do not just discover what is out there—they discover what we did not know we were even missing . Our powerful tools illuminate the deep gaps in our knowledge—they expose our ignorance. That is how many breakthroughs begin— not just by reinterpreting the old map, but by inventing a new kind of compass we had not yet imagined. So using a new tool is not just a technical upgrade, it is often actually the birth of a whole new kind of science that stretches our sensory reach. The connection between toolmaking and the new types of questions it sparks is not just a success story for science—it is a principle for discovery. Scientific progress is driven by tool innovation and planning—not just serendipity or theories (Chapters 2 and 6). We can summarise the central point here: the particular tool we apply—whether a particle collider, statistical method or brain scanner—largely determines how we can design, carry out and analyse our study. It sets how we can define and collect data. It shapes the scope of our results, what conclusions we can draw and even how other researchers can test and confirm evidence. In other words: we tailor and fit our research around the current tools we use. We need to let go of the common belief that tools support research, they largely define it: they are our lenses that determine what parts of the world come into focus and what parts do not. The biggest breakthroughs are often less about coming up with the perfect questions and more about coming up with the right tools. In short: science is the pursuit of answers to both known and unknown questions by creating and upgrading our tools of discovery—tools that expose new realms and tackle our ignorance. There is a deeper insight here: a tool is more than an object—it is a language of discovery. A tool like x-ray crystallography or electron microscopy does not just improve science in one lab—it transforms the kinds of questions we can ask and how we talk about and tackle them across entire fields. While our attention is focused on science's great discoveries and theories (final outputs), our ever-expanding toolbox is what makes all steps throughout the discovery process possible. With it, we do not just uncover knowledge—we build the process of discovery itself. Here we lay out a conceptual framework of science and discovery— what we call the new method-to-discovery view (MTD). The idea is straightforward: when we extend our toolbox, we can study and measure the world in powerful new ways that enable us to develop new knowledge and discoveries—and these make up Our adaptive toolbox (tools – from telescopes to statistics – that reduce the limits of our mind and methods) Our universal toolbox (our mind's abilities to observe, solve problems and experiment) The world (nature and society) Develop new methods to tackle the limits of existing ones in observing and measuring the world used to develop evolved Human beings Data about the world New knowledge and discoveries (evidence and theories) Science (all discoveries, methods and fields make up science – our bodies of knowledge) Medical, technological and social advances to develop cumulatively as applied to make enables us to study and measure and collect Figure 9.2 How expanding our toolbox drives new knowledge and discovery T science (see Figure 9.2). Here, a powerful feedback loop drives toolmaking: when existing tools fall short—whether hitting observational blind spots or measurement gaps—we build better tools, but over time those tools reveal limitations and demand better tools themselves, and the cycle begins again. How do methods and tools connect data to results and to theory, and make both results and theory possible? How do different methods and tools produce different evidence about the same phenomenon? This MTD view offers insights: it explains how methods guide each step of the scientific and discovery process—acting as a compass steering the path to discovery (Figure 9.3). The data in a study can only be defined, collected and analysed after we select a method—from randomised controlled trials, to CRISPR gene editing, to gravitational wave detectors. How we assess experiments and studies also depends mainly on our methodological design. Methods define standards of evidence, replicability and precision. What we can discover in science largely begins and ends with our methods and the evidence they produce (Chapter 6). What questions we can ask How we design and conduct a study How we observe, measure, and collect dataWhat causeeffect relationships we can study How we analyse a study What evidence we can create How we interpret results What theories we can formulate Methods and tools shape each step in how we do science How we define phenomena How we replicate discoveries Figure 9.3 Methods-driven science: how methods and tools shape knowledge and discovery at each step Other factors—like serendipity and theories—can influence part of the discovery process. But methods are the most systematic and universal framework that consistently shapes each step of the discovery process across breakthroughs and fields. Methods are the procedural backbone that makes discovery possible, 252 T HE ENGINE OF SCIENTIFIC DISCOVERY testable and generalisable, while other factors generally depend more on contexts or fields (Chapter 6). In contemporary science, we can only develop theories, directly or indirectly, by relying on the tools that uncover findings demanding an explanation; and tools are then what let us test whether our theories actually hold up (Chapter 6). Solid analysis depends on solid data. And solid data only ever comes from a solid method. That is why designing the right method is essential for running experiments. Evidence is only as strong as the methods used to generate and test it. When researchers using different methods reach the same results, our theories become more reliable. For example, climate change is not confirmed by just one method: it is shown independently by statistical climate models, satellite temperature measurements and ice core samples from Antarctica that record CO₂ levels. That convergence is commonly how our evidence and theories become established science.
Can we also better understand science and discovery by exploring the philosophical assumptions that underlie our methods? Indeed, each method and tool we use shapes how we are able to see the world—through five domains: • Ontology: Using a specific method, we define and categorise what we are even able to study—guided by the method's design. With statistical methods for example, we can only study what we can quantify—capturing at times complex reality in data observations. • Measurement system: Using a specific method, we set how we can represent what we observe. We can record 'yes or no' responses, rate a phenomenon on a scale, or measure exact concentrations with our data (binary, categorical or continuous variables). • Causal relationship : Using a specific method, we choose what kinds of causeand-effect relationships we can reveal. Statistical methods can uncover average causal effects of a treatment or policy (giving us probabilistic insights), while algebraic equations generally describe fixed relationships between factors (deterministic). • Level of evaluation: Using a specific method, our methodological choice fundamentally determines what aspect of a phenomenon we explore. It shapes whether we study outcomes (like medical diagnoses using clinical trials) or processes (like tracking behaviour using observational studies). It determines whether we investigate systems (like our ecology using computational modelling) or properties (like chemical substances using spectroscopes). • Epistemology: Using a specific method, we set what counts as valid evidence and explanation. The scope of our knowledge is shaped by the scope of the method we employ to create that knowledge. How rigorous our climate predictions T for example are depends on how rigorous our statistical and computational methods and satellite observations are. Each tool has these assumptions built in. Our tools standardise and automate how we do science—making it possible for different scientists to investigate the same problem in the same way. We can illustrate this with a clear example: randomised controlled trials. RCTs are widely seen as the best method in medicine and social science for assessing how effective a treatment, drug or policy actually is. When we apply the RCT method, we are studying one targeted cause—the treatment—and isolate its effect from the broader environment. This research design answers questions like 'Does this vaccine reduce infection?' but is less suited for exploring complex webs of causes or social systems. Using the RCT method, we measure the outcome (did it work)—not the process (how did it work). We focus on the average causal effect on everyone in the trial—not the range of effects across individuals. RCTs are best for simple, well-defined interventions but struggle with complex, interconnected interventions like urban planning or climate policy. These built-in features of any method do not just shape our results. They shape how we see the world itself: dictating how we define, measure, evaluate, establish causation and ultimately what we count as reliable knowledge. Compared to RCTs, methods like observational studies, in-depth case studies or network analysis let us examine similar questions but from different perspectives. While the RCT method has transformed how we fight disease and design public policy, it cannot study other aspects of the same phenomena—where other methods fill in the gaps. We cannot emphasise this enough: the way we understand the world is inseparable from how we measure it. The cause-and-effect relationships we uncover—or miss entirely—depend on the available tools we use to detect them. Our current methods are the lenses through which we see reality—the bridge between the world and how we examine and explain it. To describe how we spark discoveries is largely to describe the methods we leverage to study, measure and reveal. This methods-driven perspective also helps explain why scientists often argue: namely how major debates and disagreements in science arise. Big debates in biology—from the nature of life to how much nature or nurture shapes human behaviour—generally emerge from using different methods to study genes, our natural environment or social context. Big debates in physics—from the nature of matter to the universe—are often traced back to applying different methods. Think of deterministic algebraic equations or probabilistic statistics to express these phenomena. Big debates in medicine and economics—from the most effective treatments to the best government policies—typically arise from relying on different methods. Think of randomised trials, observational studies or network analysis. Different methods push certain aspects of a phenomenon into the foreground and others into the background. They frame the very way we think about and conceptualise the same phenomenon. After outlining how our powerful methods drive science from five different perspectives, we next expand on the general theory of science (which we introduced in Chapter 6). 254 T HE ENGINE OF SCIENTIFIC DISCOVERY