A discussion tree is a tool for illustrating and understanding discussions. It’s a tree diagram meant to outline/summarize the ideas and visually show the discussion structure (what is a reply to what, what wasn’t replied to, size of discussion branches). Trees help objectively figure out the current state of the debate/discussion and help reach a conclusion rationally.
Explaining how discussion trees work also serves as a way of explaining how rational discussion works.
All written or spoken discussion is externalized thought, and lots of internal thought is self-discussion (discussing with yourself). Consequently, understanding rational discussion is crucial to understanding rational thinking.
Creating discussion trees involves breaking complex ideas into smaller parts and specifying relationships between ideas. These are essential parts of thinking.
There are useful words for talking about trees. Each box in a discussion tree is called a node. A node can be e.g. a statement, claim, argument, explanation, question or comment.
Lines indicate relationships between nodes. Nodes 2 and 5 are parent and child (2 is the parent, 5 is the child). Use child nodes to reply to the idea/argument in the parent.
A descendant is a child, grandchild, great grandchild, and so on – any node linked by one or more child relationships. The opposite, a node linked by parent relationships, is an ancestor.
Node 1 is the root (or head) node because it has no parent. It’s the start. Nodes 4, 6, 7 and 8 are leaf nodes because they have no children. Leaves are the outside of the tree. They’re notable because they’re unanswered – there is no reply to a leaf.
A subtree (or branch) is a node and its descendants. It’s a tree starting with a different root. For large trees, it’s often useful to focus on understanding or discussing one branch at a time. A group of nodes is a partial subtree: some nodes of a subtree (and their descendants) can optionally be left out.
The level of a node is its distance from the root. The root is the topic (level 0). Node 2 is at level 1 (a 1st response to the topic) and node 5 is at level 2 (a response to a response to the topic). In two person discussions, it’s common that one person says everything on an odd numbered level and the other person says everything on an even numbered level.
A node is resolved – a conclusion about it has been reached – if the people in the discussion agree on whether it’s correct or incorrect. Mark resolutions on the tree. You can also resolve nodes according to your own opinion. You can mark on the tree the conclusions you’ve reached, the conclusions the other guy has reached, and also the conclusions you both agree on.
Consider a sentence like “I think W because X, Y and Z.” This single sentence contains four ideas. A tree could represent it with four nodes. Trees encourage people to separate their ideas so they can look at each idea by itself. Trees also let people conveniently see how ideas relate to each other. E.g. W is an argument against another idea, V (its parent), and V has several parts (children) and is an argument against U (its parent).
For breaking ideas into parts, try writing out the idea with simple sentences using no conjunctions or punctuation, then give a node to each sentence. Alternatively, write your ideas normally but then look for different parts. Conjunctions like “because”, “if”, “or”, “but” and “and” are indicators that there are multiple ideas put together.
Writing (bullet) point form notes is another way to break ideas into parts. Conveniently, point form notes (with indenting for sub-points) can be imported into mind mapping apps and automatically converted to a tree.
Multi-part ideas form a group. The head (root) of that group is the leader. It says the main point of the group, e.g. to claim W or argue against W. The other nodes in the group help support that purpose, e.g. by providing additional reasoning, info, details, explanation, answers to potential questions and links to sources (e.g. books or webpages with more info).
Tip: If you don’t already know how to break a specific idea down into parts, that means there’s room for improvement in your understanding. Try figuring it out, e.g. by brainstorming some parts it might have. When you have a clear understanding of the parts of an idea, and breaking it down would be easy, then you’re in a position to judge whether including the parts in your tree would be useful or would be unnecessary detail.
When you start understanding the parts your ideas are made of, it can be overwhelming at first. You’ll realize there’s lots of complexity in things you found “obvious”. It’ll take time and thought. But you’ll get more used to it with practice. It gets easier and faster. And then you understand your ideas and the world better and you can more easily learn new things. (A general pattern with learning new things is that, the first few times you do it, it’s a lot harder and slower. You should expect that with making discussion trees diagrams too.)
How small a part or chunk should ideas be split into?
The minimum is the smallest version of an idea which makes sense as a single thought. Can it stand on its own, independently and autonomously? This is approximately one simple sentence with one verb. English and other languages are designed with the concept of a sentence being one thought, and if you don’t use the stuff that makes sentences more complicated, then you get roughly the minimum amount of stuff for a meaningful thought.
A single noun, like “cat”, is too small. What about the cat? Similarly a verb, like “want”, is too small. Who wants something? What do they want? “My cat wants tuna.” is a pretty minimal English sentence that is enough to express a meaningful thought.
Connecting multiple independent ideas allows for creating more complex structures. Complex sentences (sentences with multiple clauses), paragraphs and whole articles do this. They build up more sophisticated concepts by combining many simpler ideas.
Similarly, ideas are made by combining simpler parts (like “tuna”) that are less than one idea. “Tuna” requires some context, some filling in the blanks with guesses, to convey a meaningful thought. E.g. if you imagine an open can of tuna sitting on a table, you’re adding information that isn’t in the word “tuna”. Tuna could be at any location, in any form, e.g. a currently living tuna fish in outer space about to fall into a black hole. But even that is making assumptions. In the sentence “I wish I had tuna” there isn’t any tuna in reality, so imagining some tuna existing in reality doesn’t fit all uses of the word “tuna”.
You can break ideas into larger parts than the minimum. In general, when there are problems, e.g. people disagreeing, break ideas down more. But when people understand and agree on some point, and don’t see a reason to break it down more, then that’s good enough and you can focus your attention elsewhere. However, people tend to underestimate how much they should break things down, and overestimate what they understand and agree with, so err on the side of breaking ideas into smaller parts than you think you should.
A minimal, indivisible part is called an “atom”. That’s what the word means even though science has discovered that atoms are actually made of sub-atomic particles like neutrons. Combining atoms gets a molecule. Combining molecules can make a cell. Combining a huge number of cells can make an organ like a heart. Combining organs and a few other things can make an animal.
You can use the atomic model as a metaphor to think about ideas and writing. A minimal idea is an atom, which is roughly a simple (one clause) sentence. A paragraph is a molecule. A section is a cell. A chapter is an organ. A book is an animal. This is how smaller parts combine to create a more complex and greater whole.
Atoms combine into molecules with bonds. There are both atoms and relationships (bonds) connecting atoms. Similarly, molecules combine with bonds. A cell has a cell wall which keeps everything together in one group with a clear division between the group (cell) and everything else (the external environment and other objects). Organs too have outer edges and have their cells attached together, and animals have skin.
Discussion trees are about, at the same time:
Another useful metaphor is that simple ideas are building blocks and our conclusions are buildings (it could be a one story building or a skyscraper depending on how complicated the idea is).
We combine building blocks in intermediate stages, e.g. we combine many rocks (and some other stuff) to make concrete. Then the concrete is used to make a pillar which is one of the supports to create the next level, which is one of the levels of the building.
Buildings are made of many little parts which are combined into bigger and bigger pieces of the building (like a wall, a room, or a whole floor). Again, there’s building up complexity from many smaller parts which are connected together into medium parts which are connected together into big parts (so the big part consists of many small parts).
Organizing thinking involves figuring out a good level of detail to use as the simplest level for what you’re doing. Examine things at that level of detail. For more complicated things, break them into parts to reach the right level of detail so you can understand them in a simpler way. Figuring out how something is a combination of simpler parts lets you focus on one part at a time to learn and understand it, and enables you to judge the correctness of the parts. Looking at simpler parts also lets you look at and understand the relationships between those parts.
If it’s not working, try considering some things at a simpler level. That’s a good way to get unstuck, and overcomplicating things (without enough understanding to deal with it) is a common weakness that people in our society have.
You can always break things down more until you get to simple sentences. And even then, you can break those into things that are less than a thought (phrases or words) and consider those individually. Sometimes discussions fail because people have incompatible understandings of a word.
Many ideas that adults deal with in their lives are very complicated. They can be broken into parts which are broken into parts that can be broken into parts, and so on, several more times, before getting to simple sentences. A common error is trying to talk about the complicated ideas without understanding the simpler ideas they’re built up from, and how those simpler ideas are organized (what’s related to what, what’s built out of what).
Tip: You can sometimes skip writing out a tree for your knowledge if you could write that tree. If you understand something well enough that a tree wouldn’t be very hard to make, then maybe you don’t need to write out the tree. But if you couldn’t create the tree, or would find it hard, then that means you’d learn a lot from the tree. Also, be careful trying to skip writing out trees until you’ve written over 100 trees – you need practice and experience to develop a good understanding of what’s skippable.
Nodes have words to explain what they mean. Relationships between ideas are represented with a line, not words. This works OK for simple relationships, especially when the words in the node give hints about the relationships. What about complex relationships?
When there’s a complex relationship that needs explaining, use a group of nodes. A counter-argument can have a head node which summarizes the argument, plus some child nodes with facts and claims, plus some other child nodes which explain how the facts and claims work together to make an argument and how that argument relates to the group’s parent that you’re trying to criticize.
Or in a simpler case, you could have a group of two nodes. The first is a claim and the second (its child) gives some reasoning explaining how the claim relates to the claim’s parent. Whenever a relationship isn’t simple enough, you can add nodes explaining.
This doesn’t just apply to critical arguments. Say you make a claim and you have five pieces of evidence related to the claim. But it’s not clear enough why three of those pieces of evidence help the claim. Then for those three, you could add a child node explaining how that evidence relates to the claim.
When you add an idea to a tree, it’s up to you to decide how many nodes you need to explain it (just one or a group). And if it’s a group, it’s up to you to decide how to organize those nodes and what to put in them. That can include facts and claims as well as explanations of why they matter and how they connect to the issue at hand.
The point is, you can use nodes to talk about the relationships between other nodes. That lets both the nodes and the relationships have written explanations. A simple way to do this is to have some nodes that say “Relationship:” at the start, or which have a specific color, that explain the relationship between the two nodes they connect. If you use that a lot, then every other level of the tree could explain relationships.
A node or group can be an argument. An argument either positively supports or negatively criticizes its parent node. (For a group, find the root node of the group, then look at its parent node.)
Decisive (also called conclusive or essential) arguments argue that the parent is incorrect. That implies that at least one of the parent or argument must be incorrect. They can’t both be correct. They’re incompatible. If a decisive argument node or group is resolved as correct, then its parent must be resolved as incorrect. The parent is refuted by that criticism.
Positive arguments, inconclusive negative arguments and explanatory comments are never decisive arguments.
Decisive arguments shouldn’t be ignored. They’re mandatory to address. Other nodes don’t necessarily have to be dealt with. You can judge a non-decisive node is uninteresting and unimportant (and explain why). But with a decisive argument, saying it’s unimportant is an inappropriate response (unless you think its parent, the thing it’s criticizing, is unimportant). An argument that a claim is incorrect is important and should be evaluated.
Figuring out which arguments are decisive or not, and focusing on making and resolving decisive arguments, is the most effective way to reach a conclusion. That’s the proper focus of debate. In the alternative, if your focus is on learning, then decisive arguments are less important and everything else, especially questions, explanations and information, has a larger role.
Marking/labelling decisive argument nodes or groups, or indecisive/inessential nodes and groups, is one of the standard ways to improve and better understand a discussion tree. You can mark with e.g. colors, words, icons or by drawing lines around groups.
A question counts as a decisive argument if not having an answer to the question would mean the idea is incorrect.
A strict debate tree is a type of discussion tree. Besides the topic (root), it contains only decisive arguments. You can convert a discussion tree to a strict debate tree by deleting all indecisive parts. More informally, you can include indecisive arguments and commentary in a debate tree as long as the decisive and indecisive parts are clearly labelled (using separate colors for decisive and indecisive parts generally works well).
In a strict debate tree, resolving the root requires resolving all nodes. Resolving a subtree requires resolving all nodes in that subtree. Even if you don’t resolve everything, you can often still resolve some subtrees. Resolving decisive arguments is the primary goal of debate.
Debate trees help you objectively evaluate the current state of the debate on some topic. What are the decisive arguments and what conclusions make sense given those arguments?
To dispute a decisive argument in a debate tree, a counter-argument is needed. To add a counter-argument, take any node in the debate tree and create a child node with a decisive criticism.
Resolving a critical argument as correct requires resolving the parent as incorrect.
When resolving a node as incorrect, check if that node is part of a group. If it’s a multi-part argument, then the whole argument is wrong if one part is wrong. So resolve the group as incorrect (some individual parts may still be correct, and could be reused in a different argument).
If all arguments against an idea are resolved as incorrect, it doesn’t mean the idea is correct. However, if no one can think of any correct or unresolved argument against an idea, then the idea should be tentatively accepted. It’s correct as far as anyone knows. People are welcome to take some time to consider, but if they can’t come up with any criticism, then they should tentatively accept the idea.
What if there are multiple incompatible ideas with no criticisms? When there are alternatives, how do we choose? If two ideas X and Y contradict each other and both fail to explain why they’re correct, then they’re both inadequate. Criticize each of them for reaching a conclusion that is more specific than their arguments allow. Idea X made X claim about reality even though its reasoning was only good enough to reach the conclusion that X or Y is how reality is. Accept instead the idea Z which says that X or Y could be correct. Z is the single known idea about this topic that has no criticism of it.
Many discussions or debates don’t reach conclusions. People bicker endlessly and the topic keeps changing. A technique that helps is to focus on resolving one subtree at a time. Once a node is brought up, only discuss descendants of that node until the node is resolved. During that discussion, when a new node is brought up, again focus only on its descendants until it’s resolved. And so on.
This is called a depth-first search because it considers higher numbered (deeper) levels first rather than other subtrees at the same level first (breadth-first). You don’t have to use depth-first every time, but it’s a useful and underrated approach. It especially helps when discussion is chaotic.
Ideas other than arguments are important and valuable. Explanations of concepts help us understand the world (and help lead to arguments). Also, if your indecisive argument is genuinely important to a debate, then you can reformulate it as a decisive argument.
A explanation, piece of information or an indecisive negative argument is useful for brainstorming and inspiration. That is important. But if what you want is a decisive negative argument, you’ve gotta come up with one. To argue more decisively, consider what the goal is (what problem is an idea trying to solve) and then what specifically is required to reach that goal, and then criticize how an idea (aka solution) won’t meet those requirements.
If your comment is a positive argument, in addition to using it as inspiration, you can translate it to a negative argument. All correct positive arguments can be translated to negative arguments.
Here’s how to translate: A positive argument points out a positive feature of an idea. To get a negative argument, point out how one or more alternative ideas lack that positive feature. For this argument to be decisive, explain how lacking that positive feature makes the idea incorrect (the idea needs that positive feature to work).
In all cases, you must come up with a decisive negative argument to change the objective state of the debate and contribute to a debate tree. You also must come up with a decisive negative argument to refute anything in any discussion tree.
Note: People often use positive arguments in informal discussions without translating them to negative arguments. That’s OK as long as long as everyone understands and agrees. It’s a shortcut. But if anyone disagrees or objects, then the standard way to resolve that disagreement should be to translate the positive arguments to negative arguments. Don’t insist on using positive arguments when there’s a dispute.
Also, positive arguments are often OK, especially informally, internally within groups. The group type (decisive argument or not) is determined by the group’s root node. Positive reasoning can help explain a negative argument and break it into parts.
Sometimes people agree to resolve a node as correct or incorrect, but then later they think of a new idea or argument and want to make the node unresolved again. This is OK. Disallowing it would discourage anyone from agreeing to resolve a node.
Rather than leaving everything unresolved forever just in case we think of a new argument later (possibly based on new evidence), we tentatively resolve nodes. We use our best judgment to reach conclusions when we think we know enough. But we’re allowed to reconsider later. Judgments don’t have to be final.
Discussions, including debates, should try to seek the truth. Discussion trees help organize and keep track of discussions, which helps figure out what’s true. Trees help show which ideas are answered or unanswered.
Answered ideas have children which express arguments, doubts, objections, etc. Unanswered ideas have no children, or have only children which have been resolved as incorrect.
Trees are also useful when discussions get confused or chaotic. I can make a discussion tree to share my understanding of the discussion, and you can point out which parts you think don’t match the discussion we had. Or we can both make trees and compare them for differences.
We can discuss by adding nodes to a tree as we go along so we’re keeping the discussion clear and organized the whole time. Or we can create a tree and start adding nodes if a discussion has trouble.
Regardless of whether you agree or disagree with an argument, present it fairly. This is easier with trees than normal because you’re usually summarizing and can refer to sources like books or articles. You don’t have to personally present all the details of arguments that you disagree with.
Each side of the debate is an intellectual position (or school of thought) – a group of related ideas that fit together. Don’t associate the sides too strongly with people. People can try to contribute arguments for every position. Everyone’s goal should be to evaluate the ideas and try to figure out a good conclusion. The point of discussion is to help each other do this.
It often helps to color-code or otherwise label nodes based on which side they’re on. This makes argument groupings clearer (multiple connected nodes of the same color are a multi-part argument) and it makes it easier to see criticisms and objections (a child of a different color).
Keep in mind there are usually more than two sides. It’s often helpful to focus on two sides at a time, which keeps things simpler, but remember that there may be other viewpoints. If you finish a tree with two sides, consider if there is another position on the same issue to investigate next.
You can make discussion trees by yourself. You can use them to help study discussions between other people. E.g. you could read arguments about abortion and put them into a tree. Figure out what’s a decisive argument and what’s a comment. Figure out what’s an unanswered leaf node. Consider which nodes you’re willing to reach a judgment about and resolve. After you begin to understand other people’s arguments, add some nodes and comments of your own. Those activities will help you think and learn.
Creating a discussion tree is also a good way to express your understanding of a topic. You can write all sides of the discussion. Write the important arguments and questions for the other side that you know of, and give your responses. Then you can show the tree to people and ask what you’re missing, rather than starting a discussion from scratch.
You can also use a discussion tree to help figure out a topic by arguing with yourself. If you don’t have a conclusion yet, you can use the tree to express what you do know and work towards reaching a conclusion.
In general, if you can’t make a tree and reach a conclusion alone, you shouldn’t be debating others. You don’t know what you think so why try to correct other people and argue about what they should think? If you don’t already have a conclusive tree, in your opinion, then you shouldn’t do advocacy. Stick to discussions where you ask questions and share ideas.
You can make trees with different levels of detail. Usually it works best to summarize arguments but leave out most of the details. This helps you see an overview of the discussion and its structure. People can refer to source materials for details. E.g. if you have a debate on a web forum, people can look at the forum posts for more details about arguments. Or if your tree relates to some books and articles, people can get the full details from those.
You can write out more detailed arguments in a tree when you find it useful. Sometimes I write a long paragraph in one node. But I don’t find it useful to put a whole 1000 word essay in a node. I’d rather just give the link to it. Alternatively, I’d break the ideas in the essay up into separate pieces and use them for multiple nodes.
Figuring out how to break ideas down into smaller pieces (that still make sense individually) – one node per piece – helps understand ideas better.
Making a discussion tree is a similar activity to outlining. It has many of the same benefits. E.g. you can make trees (or outlines) about a book or chapter you read. That helps you e.g. remember it and understand how the different ideas fit together.
When trees get too big, you can put subtrees into separate documents. You can make an overall tree where each node is the name of a subtree. You can also make a mixed tree with some regular nodes and some nodes which refer to subtrees in other documents.
Referring to a tree in another document is like citing a source. Having a group of arguments as a tree in a document lets you reuse them in many future discussions.
Referring to a tree from another document is equivalent to adding all its nodes where the reference is. It’s as if you copied over every node one by one. It’s a shortcut.
Pro/con lists are a common tool for people trying to organize the arguments about an issue. Some biased people only care about arguments for “their side”, so they wouldn’t use a pro/con list because one half would be blank. Using a pro/con list is an improvement over that which helps people be more objective and consider both sides.
Trees have advantages over pro/con lists because they break ideas into parts and show relationships between ideas. Trees show counter-arguments and counter-counter-arguments. Pro/con lists encourage focusing on level 1 arguments.
Pro/con lists don’t help people break their reasoning into smaller, related parts to better understand their thinking. Each point gets one statement and that’s it.
Pro/con lists are partly based on mistaken philosophy ideas about how rational debate works. The idea is to add up the weight of the arguments for each side, subtract the weight of the arguments against that side, and whichever side has the best score wins. The hidden assumption here is that most arguments are inconclusive, so for complex issues there will usually be a bunch of inconclusive arguments on each side. So people try to evaluate how good the inconclusive arguments are and sum them together.
The correct approach to debate is to focus on decisive arguments. Come up with reasons that things can’t be right. An inconclusive argument against X is an argument which is compatible with X being true. Having a bunch of those isn’t good enough. A conclusive argument is one which is adequate to reach a conclusion about some issue (e.g. that a particular claim is mistaken), given what we know now (things could change if with new information or arguments). An inconclusive argument is (by definition) inadequate to reach a conclusion, and having ten of them can’t fix that problem. If multiple inconclusive arguments could work together to reach a conclusion, then what you should do is make a multi-part, conclusive argument which uses them.
Debate trees fit the concept of arguments conclusively refuting other arguments. They focus the discussion on critical thinking. They encourage trying to find mistakes in the ideas (conclusive problems, not irrelevant complaints about minor details).
Pro/con lists aren’t designed for resolving nodes. Debate trees are better at reaching conclusions. Debate trees are also capable of reaching conclusions about subtrees, which is an easier way to get started and is partial progress, while pro/con lists are too all-or-nothing and only focus on reaching a final conclusion all at once.
Note: These apps can make discussion trees with nodes going in multiple directions from the root node. That’s just an optional visual change. It doesn’t affect the tree structure (the parent/child relationships) compared to a normal looking tree where all descendants are below their parent (or to the right for a horizontal tree).
View my idea trees blog category.
I explain the importance and philosophy of decisive arguments in Yes or No Philosophy, which is also one of the first places I talked about trees. I also talked about decisive arguments in my Critical Fallibilism Course.
You can learn more about using idea trees in my Tutoring Max video series. In those videos, I helped teach Max about trees, and suggested practice activities that would be useful for most people. There’s more introductory material about trees in Tutoring InternetRules.
I originally wrote this article in late 2019. I did a light editing pass and posted this version in June 2021.