What I want to hear during a scientific presentation

Lorenzo Clemente

University of Warsaw

​

LMW@LICS’25 @ Singapore, 23/06/2025

Disclaimer

• What works for me needs not work for others
• Problems are more important than solutions
• Take what you like, discard the rest
• I do not believe all that I say
• This presentation will contradict itself
• We are all learning

You are the expert

You are the best person in the world to do research in your area

• You know all about the topic
• You can quickly get to the relevant technicalities
• You can quickly solve the easy problems

​

You are the worst person in the world to explain your area

• You forgot that it took quite some time for you to become an expert
• You cannot expect the audience to follow your journey in a short time
• You forgot the difficulties you had when mastering the topic

You are the expert

You are the best person in the world to do research in your area

• You know all about the topic
• You can quickly get to the relevant technicalities
• You can quickly solve the easy problems

​

You are the worst person in the world to explain your area

• You forgot that it took quite some time for you to become an expert
• You cannot expect the audience to follow your journey in a short time
• You forgot the difficulties you had when mastering the topic

Anecdote

• At the beginning of my PhD I was a TA for functional programming course
• I never programmed in Haskell before
• I was learning the material in parallel with undergrads
• Best teaching experience ever:
• I knew exactly what were the key difficulties in learning the topic

Take the point of view of the audience

What is the target of your presentation?

​

• Local seminar
• most of the audience will not follow you
• you have some time to lecture some basics
• Specialised technical workshop
• most of the audience will follow you on the more technical part
• International conference
• nobody will follow you
• the audience will be tired, distracted, answering emails
• can’t afford technicalities

Take the point of view of the audience

Conflicting objectives

​

• You
• Want to show your great technical achievement.

​

• Me (the audience)
• Won’t be able to digest technical content in a short time.
• Want to connect your achievement with what I know already.
• Want to know what is new vs. what is known.
• Want to know what is easy vs. what is difficult.
• Want to know why the result is relevant.
• Want to know how can the result be applied.

Take the point of view of the audience

Conflicting objectives

​

• You
• Want to show your great technical achievement.

​

• Me (the audience)
• Won’t be able to digest technical content in a short time.
• Want to connect your achievement with what I know already.
• Want to know what is new vs. what is known.
• Want to know what is easy vs. what is difficult.
• Want to know why the result is relevant.
• Want to know how can the result be applied.

How learning works

Learning is about connecting different concepts.

• Present new concepts by connecting them to common background.
• Minimise the amount of new information.
• If unsure whether something should be included or not, remove it.
• Skip formal definitions, but say how it relates to the rest.
• BUT! No shortcuts
• Better to avoid a semi-formal statement if it will confuse people
• You will be punished by a lot of questions

Strategies

Place your result in the bigger picture

• Refer to previous works
• helps the experts to categorise your result
• acknowledges prior work
• likely some member of the audience contributed to the topic

​

Stress what are the difficult/non-trivial steps

• non-experts need to be told what is easy and what is not

Strategies

Place your result in the bigger picture

• Refer to previous works
• helps the experts to categorise your result
• acknowledges prior work
• likely some member of the audience contributed to the topic

​

Stress what are the difficult/non-trivial steps

• non-experts need to be told what is easy and what is not

Strategies

Avoid formal definitions

• Focus on what a definition achieves rather than on how it achieves it

​

Technical concepts best explained

• in analogy to well-known concepts
• by simple examples

BUT!

• good examples are difficult to construct
• it may not be clear what are the key ingredients
• don’t get carried away too much with complicated stories

Strategies

Avoid formal definitions

• Focus on what a definition achieves rather than on how it achieves it

​

Technical concepts best explained

• in analogy to well-known concepts
• by simple examples

BUT!

• good examples are difficult to construct
• it may not be clear what are the key ingredients
• don’t get carried away too much with complicated stories

Progressive escalation of content

1. Everybody knows it
2. Somebody knows it
3. Nobody knows it

​

Idea: Give everybody something to take away from the presentation

​

(A) => (B): Everybody should learn something new

(B) => (C): Somebody (the experts) should learn something new

An example

• Everybody knows it

Let a series f_1 be regular if there are generators f_2, …, f_k s.t. for every input symbol a,

the left derivative of generator f_i by a belongs to {f_1, …, f_k}

(this is in analogy to DFA and regular languages)

​

• Somebody knows it

Let a series f_1 be rational if there are generators f_2, …, f_k s.t. for every input symbol a,

the left derivative of generator f_i by a belongs to the linear span of {f_1, …, f_k}

(this introduces the new notion of linear span)

​

• Nobody knows it

Let a series f_1 be shuffle-finite if there are generators f_2, …, f_k s.t. for every input symbol a,

the left derivative of generator f_i by a belongs to the algebra generated by {f_1, …, f_k}

(this introduces the new notion of algebra)

An example

• Everybody knows it

Let a series f_1 be regular if there are generators f_2, …, f_k s.t. for every input symbol a,

the left derivative of generator f_i by a belongs to {f_1, …, f_k}

(this is in analogy to DFA and regular languages)

​

• Somebody knows it

Let a series f_1 be rational if there are generators f_2, …, f_k s.t. for every input symbol a,

the left derivative of generator f_i by a belongs to the linear span of {f_1, …, f_k}

(this introduces the new notion of linear span)

​

• Nobody knows it

Let a series f_1 be shuffle-finite if there are generators f_2, …, f_k s.t. for every input symbol a,

the left derivative of generator f_i by a belongs to the algebra generated by {f_1, …, f_k}

(this introduces the new notion of algebra)

An example

• Everybody knows it

Let a series f_1 be regular if there are generators f_2, …, f_k s.t. for every input symbol a,

the left derivative of generator f_i by a belongs to {f_1, …, f_k}

(this is in analogy to DFA and regular languages)

​

• Somebody knows it

Let a series f_1 be rational if there are generators f_2, …, f_k s.t. for every input symbol a,

the left derivative of generator f_i by a belongs to the linear span of {f_1, …, f_k}

(this introduces the new notion of linear span)

​

• Nobody knows it

Let a series f_1 be shuffle-finite if there are generators f_2, …, f_k s.t. for every input symbol a,

the left derivative of generator f_i by a belongs to the algebra generated by {f_1, …, f_k}

(this introduces the new notion of algebra)

Presenting other people’s work

Some of the best presentation I’ve seen were about the work of other people

• A member of the audience delivers the presentation!

=> They focus by construction on what the audience wants to hear

• They avoid the common pitfalls
• They are not technical
• They are easy to follow
• cognitive bottleneck of the speaker
• They focus on what is known already

​

Idea: Conference where we present each other’s results?