What is scientific creativity—and how do you feed it? (Part I)

acid-citric-citrus-997725

Last winter, on a speaking trip to Norrköping, someone asked me to write about skills (and meta-skills) that scientists and PhD students need, beyond writing papers. Turns out that this is a lot more difficult than writing about writing, where the end product—a scientific paper—is something tangible and amenable to analysis: how do great introductions look like? How do the greatest writers finish their papers? It is much more difficult to write, say, about learning to be creative, which is what I shall try to do here. But what would be more important for aspiring scientists than creativity?

Science is all about creativity: coming up with the right questions, developing clever methods to answer those questions, and connecting the answers in imaginative ways to learn something greater. But we rarely talk about creativity as a skill—often, people view it as something that you either have or don’t have, just like an ear for music or an eye for design. And just like with music and design, this view is wrong: everything can be learned. So how do you learn to be creative?

Before attempting to answer this question, let’s take the bull by the horns and ask what creativity is. If by creativity we mean the ability to bring forth ideas that are entirely new, we are immediately hit by a very difficult, philosophical question: where do new ideas come from? At least to us (recovering ex-) physicists, the emergence of something that wasn’t there before is kind of strange: aren’t there conservation laws that forbid this kind of travesty from happening? What is it that gives birth to new information (because that is what happens when a new idea emerges, whether it is a question or an answer)?

If physics doesn’t provide us with answers, let’s drop it for a while and put on the hat of a biologist: in the realm of living things, don’t new things gradually emerge, driven by the slow Darwinian evolution? Notice the word “gradually”—biological evolution is slow tinkering, a process where existing forms and shapes and organs are gradually transformed into something new, of dinosaurs developing feathers that eventually help some of them to learn to fly, of finches’ beak shapes adapting to their habitats. So in biological evolution, everything that is “new” is built on top of a lot of something old, and this happens slowly: a slow expansion into the adjacent possible, if you’ve read your Kauffman.

Are there some other natural processes where new forms emerge more rapidly? The human immune system provides a great example. Somewhat surprisingly, not all our cells carry the same sets of genes: the T and B cells of our immune system, our ultimate smart weapons against viruses and other invaders, display an enormous diversity of different receptors that recognise those invaders. This diversity results from those cells carrying some randomised (but not too randomised) parts of our genome. The precursor cells that eventually become T and B cells have strings of different modules in their genetic code, and in the process of randomisation, some of those modules are randomly picked and joined together (the rest are discarded). Then, a bit of extra randomness (extra letters, deleted letters, and so on) is added to their junction. So to arrive at new kinds of receptors, our bodies randomly merge things that are known to work (those receptor modules) and then add some noise on top. Again, “new” equals “old, but with added something.”

Let’s now return back to creativity, in the context of science or otherwise. The above examples point out that the old rhyme—“something old, something new, something borrowed, something blue”—is scientifically highly accurate, except for the blue bit perhaps. In other words, the things that we think are new are in fact modifications and clever combinations of old things, with perhaps some small amount of additional randomness. Ideas do not live in a vacuum, they emerge because of other ideas.

Therefore, creativity is the ability to merge existing ideas in new ways (while possibly adding a magic ingredient on top).

This brings us to a fairly simple recipe for feeding one’s creativity: collect lots of things that can be combined/transmogrified into something new, and then just combine them! In other words, first, feed your head with lots of information—and not just any information, but preferably pieces of information that haven’t yet been combined.

To maximise the chance of something entirely new emerging out of this process, your input information—the stuff that you feed your head with—should be diverse enough. There are, however, different possibilities: on the one hand, if you know everything that there is to know about your field, you can probably see where the holes are and combine bits of your knowledge in order to fill them. On the other hand, if you know enough about a lot of fields, you might be able to spot connections between them (think of, say, network neuroscience, applying network theory to problems of neuroscience). There are different styles here, but even if you choose to go deep instead of wide, do keep the diversity of input information in mind: just for fun, learn some mathematical techniques that people do not (yet) use in your field! You never know, those might turn out to be useful later.

To be continued…

Functional brain networks: the problem of node definition

Summary: Nodes in brain networks from fMRI are usually defined using ROI’s (Regions of Interest) so that each ROI node has a time series that is the average of the BOLD time series of the ROI’s voxels and links represent correlations between nodes. Here, we show that this averaging of voxel time series is problematic.

The human brain is a complex network of neurons. The problem is that there are about 10^12 of them with ~10^5 outgoing connections each; mapping out a network of this scale is not possible. Therefore, one needs to zoom out and look at the coarse-grained picture. This coarse-grained picture can be anatomical – a map of the large-scale wiring diagram between parts of the brain – or functional, indicating which parts of the brain tend to become active together under a given task.

But how should this coarse-graining be done in practice? How to define the nodes of a brain network –– what should brain nodes represent? In functional magnetic resonance imaging (fMRI), the highest level of detail is determined by the imaging technology. In a fMRI experiment, subjects are put inside a scanner that measures the dynamics of blood oxygenation in a 3D representation of the brain, divided into around 10,000 volume elements (voxels). Blood oxygenation is thought to correlate with the level of neural activity in the area. As each voxel contains about 5.5 million neurons, the network of voxels is significantly smaller than the network of neurons. However, it is still too large for many analysis tasks, and further coarse-graining is needed.

A typical way in the fMRI community is to group voxels into larger brain regions that are for historical reasons known as Regions of Interest (ROIs). This can be done in many ways, and there are many pre-defined maps (“brain atlases”) that define ROIs; these maps are based on anatomy, histology, or data-driven methods. It is common to use ROIs as the nodes of a brain functional network. The first step in constructing the brain network is to assign to each ROI a time series that is the average of the time series of its voxels measured in the imaging experiment. Then, to get the links, similarities between the ROI time series are calculated, usually with the Pearson correlation coefficient. The correlation between the two ROIs becomes their link weight. Often, only the strongest correlations are retained, and weak links are pruned from the network.

If the ROI approach is to work, the ROIs should be functionally homogeneous: their underlying voxels should behave approximately similarly. Otherwise, it is not clear what the brain network represents. Because this assumption hasn’t really been tested properly and because it is fundamentally important, we recently set out to explore whether it really holds.

We used resting-state data – data recorded with subjects who are just resting in the scanner, instructed to do nothing – to construct functional ROI-level networks based on some available atlases. We defined a measure of ROI consistency that has a value of one if all the voxels that make up the ROI have identical time series (making the ROI functionally homogeneous, which is good), and a value of zero if the voxels do not correlate at all (making that ROI a bad idea, in general).

Distribution of consistency for ROIs as brain network nodes
[Figure from our paper in Network Neuroscience]

We found that consistency varied broadly between ROIs. While a few ROIs were quite consistent (values around 0.6), many were not (values around 0.2).  There were many low-consistency ROIs in three commonly used brain atlases.

From the viewpoint of network analysis, the existence of many low-consistency ROIs is a bit alarming.  We also observed strong links between low-consistency ROIs – how should this be interpreted? These links may be an artefact, as they disappear if we look at the voxel-level signals. This means that the source of the problem is probably the averaging of voxel signals into ROI time series. While this averaging can reduce noise, it can also remove the signal: at one extreme, if one subpopulation of voxels goes up while another goes down, the average signal is flat. More generally, if a ROI consists of many functionally different subareas, their average signal is not necessarily representative of anything.

In conclusion, we would recommend being careful with functional brain networks constructed using ROIs; at least, it would be good to go back to the voxel-level data to verify that the obtained results are indeed meaningful.

For details, see our recent paper in Network Neuroscience.

This post was co-written by Onerva Korhonen, Enrico Glerean & Jari Saramäki.

[PS: The definition of brain network nodes is not the only complicated issue in the study of functional brain networks. Even before one has to worry about node selection, a possible distortion has already taken place: preprocessing of the measurement data. We’ll continue this story soon.]

Why can writing a paper be such a pain?

This is the first in a series of “self-help” posts for PhD students on how to write a scientific paper.

Writing a scientific paper

Show me a researcher who has never struggled with writing, and I’ll show you someone who hasn’t written anything, or who doesn’t care about the quality of the output. Science is hard, and so is writing. Together they are harder. Now add in lack of experience as a researcher and as a writer, together with the usual time pressure, and it’s no wonder that the blank document in front of you looks like the north face of Mount Everest. We’ve all been there, staring at that wall.

While no mountaineer would risk climbing Everest without a route plan, an inexperienced writer tends to neglect the importance of planning. Having no plan, she tries to do everything at once. She opens the blank document in her editor, stares at it, tries to decide what to make of her results, what the first sentence of the first paragraph should be, what the point of the first paragraph should be, and what the point of the whole paper should be.

It’s no wonder that this feels impossible. No-one can solve that many problems in parallel. Problems are best solved one at a time.

Writing becomes easier if one separates the process of thinking from the process of writing. To write clearly is to think clearly, and thinking precedes writing. Writing becomes a lot less of a struggle when you think through the right things in the right order, before putting down a single word.  A successful software project begins with the big picture: what functions and classes are needed, and for what purpose. It doesn’t begin with developing code for the internal bits of these functions and classes. A writing project should also begin with addressing the overall point and structure of the paper, before moving to details such as words or sentences.

Another way of looking at the problem is linearity versus modularity. The fear of the blank page arises out of linearity: the feeling that the only way to fill the page is to start with the first word and proceed towards the last, word by word. This is not so. Whereas reading is usually linear, writing doesn’t have to be. The process of writing should be modular – first, sculpt your raw materials into rough blocks that form your text, and then start working on the blocks, filling in more and more details, so that entire sentences only begin appearing towards the end of this process.

The approach I try to teach my students is splitting the writing process into a series of hierarchical tasks. This way, getting from a pile of results to a polished research paper is a bit less painful.

This approach begins by identifying the key point of the paper and then moving on to structuring the material that supports this point into a storyline. This storyline is then condensed into the abstract of the paper. My advice is to always write the abstract first, not last! This serves as an acid test: if you cannot do it, you haven’t developed your storyline enough.

After that, there are many steps to be taken before writing any more complete sentences: planning the order of presentation, including figures, and for each section of the paper, mapping the arc of the storyline into paragraphs so that the point addressed by each of the paragraphs is decided in advance. Then, the paragraph contents are expanded into rough sketches, and these sketches are finally transformed into whole sentences. At this point, there is no fear of the blank page, because there are no blank pages: for each section, for each paragraph, there is a map, a route plan, and the only decision that is needed is how to best transform that plan into series of words. Often, this feels almost effortless.

[Next in the series on how to write a scientific paper: how to write a great abstract]

There is now an ebook based on this series, available from a number of stores (Kindle Store, Apple Books, Kobo, Tolino, etc!)

Mitä matkapuheluidemme ajoitukset kertovat meistä

[This post is in Finnish in case you are wondering; the original English-language version can be found here. The rest of the posts in this blog are in English.]

Tämä postaus on tarkoitettu taustamateriaaliksi tiedetoimittajille, liittyen Akatemian tiedeaamiaiseen 27.4. Mutta sinun ei toki tarvitse olla toimittaja lukeaksesi eteenpäin!

Tutkimusryhmäni on tutkinut matkapuhelindataa yli vuosikymmenen ajan. Se, miksi tutkimustamme kutsutaan, on muuttunut tällä välin verkostoanalyysistä datatieteeksi ja laskennalliseksi ihmistieteeksi. Miksi tahansa sitä kutsutaankin, tutkimuksessamme tarkastellaan ihmisten käyttäytymistä laskennallisin keinoin, ja aineistot voivat sisältää jopa miljoonia henkilöitä!

Käytämme automaattisesti kerättyä, anonymisoitua, aikaleimattua dataa, joka on peräisin teleoperaattoreiden laskutusjärjestelmistä. Tämän lisäksi tutkimme dataa joka on kerätty vapaaehtoisilta koehenkilöiltä esimerkiksi älypuhelinapplikaatioilla. Matkapuhelintietojen (kuka soitti kenelle ja milloin) avulla voimme rekonstruoida sosiaalisten verkostojen kytköksiä ja tarkastella myös puheluiden aikasarjoja. Nämä aikasarjat ovat osoittautuneet erittäin mielenkiintoisiksi!

Tarkastellaan ensin hyvin lyhyitä aikaskaaloja, sekunneista minuutteihin. Jos katsomme yksittäisen henkilön puheluita, ja piirrämme aikajanalle viivan aina kun henkilö puhuu puhelimessa, saamme tällaisen kuvan:

Puheluiden purskeisuus

Tämä aikasarja on purskeinen – se on satunnainen mutta ei tasaisen satunnainen! Se sisältää hyvin lyhyillä aikavälillä tapahtuvien puheluiden purskeita (kymmenistä sekunneista pariin minuuttiin), ja pidempiä taukoja näiden purskeiden välillä. Ihmisten viestintä ja muukin toiminta on usein purskeista – eikä kukaan oikeastaan tiedä, miksi. Muuten, hermosolujen laukomisen aikasarjat näyttävät varsin samanlaisilta! Ehkä me kaikki olemme vain hermosoluja koko maailman kattavassa sosiaalisessa verkostossa… no, jätetään tämä tieteiskirjailijoille.

Mennäänpä kohti pidempiä ajanjaksoja, tunteja ja päiviä. Sieltä löydämme vuorokausirytmit, jotka ymmärretään huomattavasti paremmin. Meidän päivittäinen toimintamme seuraa päivän ja yön vaihtelua 24 tunnin jaksoissa. Poimitaanpa pari henkilöä datasta ja katsotaan, paljonko he soittavat puheluita kuhunkin kellonaikaan:

Puheluiden vuorokausirytmejä
Tästä nähdään että vaikka ihmiset yleensä nukkuvat yöllä ja valvovat päivällä, vuorokausirytmeissä on silti selkeitä eroja, mikä näkyy myös puheluiden määrässä. On aamuvirkkuja, jotka soittavat puheluita jo toisten nukkuessa, ja iltaihmisiä jotka soittelevat myöhään illalla (varmaankin toisille iltaihmisille). Me olemme kaikki erilaisia!

Vuorokausirytmeihin liittyy muutakin kuin puhelumäärien vaihtelu: esimerkiksi iltaisin puhelut kohdistuvat usein harvoille (ja läheisille) ystäville, ja päivällä ne ovat satunnaisempia.

Siirrytäänpä sitten kohti vielä pidempiä ajanjaksoja – kuukausia ja vuosia. Nyt yksittäisten puheluiden tarkoilla ajoituksilla ei ole enää väliä. Lasketaan siis koehenkilöllemme, montako soittoa hän tekee kullekin ystävistään (ja sukulaisistaan), ja katsotaan miten tämä kuvio muuttuu ajassa! Saadaan tämäntapainen kuvio:

Egosentrinen verkosto

Tämä jakauma kertoo mikä osuus henkilön puheluista suunnataan tämän eniten puheluita saavalle ystävälle, mikä toiseksi eniten, jne. Eli se vastaa kysymykseen kuinka suosittu suosituin ystävä on, ja kuinka tasa-arvoisesti me ystäviämme kohtelemme (yleensä varsin epätasa-arvoisesti, kolme suosituinta voi saada yli puolet puheluista!) Tämä heijastaa tapaa, jolla rakennamme sosiaalisen maailmamme: meillä on vain muutama hyvin läheinen ystävä ja paljon ystäviä jotka eivät kuulu tähän rajattuun sisäpiiriin. Suurin osa siteistämme on heikkoja, ja ne muutamat vahvat siteet ovat hyvin merkityksellisiä.

Tällaiset puheluiden jakaumat ovat hieman erilaisia kaikille, ja ne ovat osoittautuneet hyvin pysyviksi silloikin, kun verkostossa on suurta vaihtuvuutta. Jos tapanasi on keskittyä 1-2 läheiseen ystävään, tulet tekemään näin silloinkin, jos nämä ystävät korvautuvat joillakin muilla vaikkapa paikkakunnalta muuton takia. Vastaavasti jos jaat aikasi tasan ystäviesi kesken, teet varmaan näin jatkossakin.

Puhelujakaumilla sekä verkoston vaihtuvuudella on yhteys luonteenpiirteisiin; jos tämä kiinnostaa, kollegani Simone Centellegher on kirjoittanut blogipostauksen aihepiiristä äsken julkaistun artikkelimme pohjalta.

Onko tästä kaikesta tiedosta sitten muutakin hyötyä kuin että se on mielenkiintoista? Todennäköisesti. Käyttäjästä kerättyyn dataan perustuvat hyvinvointisovellukset ovat yksi mahdollisuus, kunhan niiden toiminta varmennetaan tieteellisesti. Tutkimusryhmälläni onkin käynnissä Helsingin yliopiston Psykiatrian osaston kanssa pilottihanke, jossa pyritään löytämään mielialapotilaiden hyvinvointia ennustavia tekijöitä sovellusten keräämästä datavirrasta.

Lopuksi vielä linkkejä alkuperäisiin tieteellisiin julkaisuihin:

  • Small But Slow World [Phys. Rev. E | arXiv] (2011)
  • Daily Rhythms in Mobile Telephone Communication [PLoS One] (2015)
  • Persistence of Social Signatures in Human Communication [PNAS | arXiv] (2014)
  • Personality Traits and Ego-Network Dynamics [PLoS One] (2017)
  • Effects of time window size and placement on the structure of an aggregated communication network [EPJ Data Science] (2012)
  • From Seconds to Months: the Multi-scale Dynamics of Mobile Telephone Calls [EPJB | arXiv] (2015)

Ant supercolonies: networks of nests

An ant (F. Aquilonia)

Ant colonies are complex systems par excellence. It’s almost as if the colony is the organism, not the ant. Ants follow simple behavioural patterns, depositing pheromones as they go and following trails of scent laid down by others. Because of their collective actions, the colony seems to have a life of its own, sprouting its foraging trails towards food sources much like a slime mold grows its branches along the shortest path to food. The colony appears to have its own reproductive cycle too: queens and males mate during the nuptial flight, and the impregnated queens then land to give birth to new colonies, like fertilized eggs. Ordinary workers play no role in reproduction; they are outside the germline.

But some species of ants behave in ways that are even more complex: they form supercolonies, networks of interconnected nests with hundreds of reproductive queens. In these supercolonies, queens and workers move freely between nests without eliciting aggression; they cooperate across nest boundaries. Ant supercolonies are the largest cooperative units known in nature: for some ants, they can extend for hundreds of kilometres.  They are also among the strangest: their existence is difficult to explain from the point of view of gene-centric evolutionary theory. This has to do with altruism: relatedness among nestmates can be low, and workers will end up helping unrelated individuals that carry a different set of genes. It may even be that ant supercolonies represent an evolutionary dead end.

Recently, I had a chance to have some fun with the genetics of ant supercolonies. My colleagues Eva Schultner and Heikki Helanterä who work on ants had collected a number of samples from tens of nests of F. Aquilonia in southern Finland. As Eva and Heikki wanted to understand the genetic structure of F. Aquilonia supercolonies, the sampled ants were genotyped for estimating genetic similarities between the nests (for technical details, scroll down). From a network-science point of view, the nests and their similarities span a weighted spatial network: nests are nodes and pairwise genetic similarities are mapped to link weights. The resulting similarity network looks like this:

2016_MY_LA_new

There are two supercolonies, one to the NE and one to the SW – the link weights inside the colonies are higher than between them, much like you would have for two communities in a social network. A closer look inside these two supercolonies (with methods more advanced than bare-bones network thresholding) revealed that there is a faint hint of substructure, of subclusters inside supercolonies. And because queens, workers, and pupae were genotyped separately and sampled at two time points, we could see that the genetic relationships between nests are not the same in terms of queens as they are in terms of workers, and not the same in spring as they are in summer when workers have started migrating.

This means that there may an extra layer of complexity in the genetics of ant supercolonies – fine structure in time and space, and in terms of class.

This work was published in Molecular Ecology last year. If you are interested in toying around with ant genetics, the data are available on Datadryad and my Python scripts can be found here: github.com/jsaramak/ants.

[Technical details: the ants were sequenced at 8 polymorphic microsatellite loci; microsatellites are nonsensical bits of DNA where a random sequence is repeated 5-50 times. They do not do anything and there is no selection pressure, and therefore microsatellite alleles are great for just seeing how close or far two populations are genetically. There are various measures for quantifying this: the simplest would be to see how often the same alleles appear in populations. In social-insect studies, the typical measure is the so-called relatedness (Queller & Goodnight 1989) and we used it in this work.]

A Neuroscience Conference On Twitter

Brain Twitter Conference ad

My colleagues at the Department of Neuroscience and Biomedical Engineering at Aalto University are organizing the Brain Twitter Conference. It takes place on Twitter on the 20th of April, with an impressive list of speakers. Talks and keynotes will be delivered under the #brainTC hashtag.

While the idea of a Twitter conference may sound like a gimmick, it should be taken seriously – not as a substitute but as something new. There are no coffee breaks or conference dinners for socializing, but anyone can attend for free. And, even better, because the tweets will remain available, a kind of time travel becomes possible – one can revisit any talk, any discussion, and any debate at will. The conference becomes frozen in time.