CHEN-BO ZHONG on information structure and creativity
MANY DAY-TO-DAY workplace activities are characterized by ‘hierarchical structure’. For example, the lean-manufacturing system categorizes all components involved into clearly-defined categories, so that workers can easily distinguish between items and use the correct components on manufacturing lines.
In many cases, information is also highly structured in the workplace. Since employees are clustered around jobs and roles, both explicit information (as compiled in a job manual) and implicit information (which is implied or understood by the worker) are categorized by job function.
Without disputing the recognized benefits of hierarchical structure, I recently conducted research with Rotman PHD Candidate Yeun Joon Kim to determine if such structure might also come with a high cost: reduced creativity.
For our work, we defined creativity as ‘combinations of information that are both novel and useful’. We used the term information broadly, per the definition of ‘declarative information’, which includes ‘chunks’ of information such as objects, symbols and facts that possess distinguishable attributes. For example, a chair contains at least three pieces of declarative information: a seat, legs and a back — each of which refers to a specific part with unique attributes that are distinguishable from the others. In this sense, in the realm of production, raw materials can be considered to be declarative information.
We suspected that a hierarchical structure featuring declarative information might be a double-edged sword: On the one hand, as research shows, it clearly increases efficiency; on the other, we felt that it may reduce the generation of creative ideas, because the presence of higher-order categories could reduce ‘distal’ or uncommon associations. For example, the inventor of the wheelchair needed to connect two distal pieces of declarative information: ‘wheel,’ which typically belongs to the vehicle category, and ‘chair,’ which belongs to the furniture category. We posited that such associations are less likely to take place when the information provided to workers is highly structured.
Hierarchichal vs. Flat Information Structures
‘Information structure’ refers to the way in which units of information are associated with one another within an information set, and it can be either hierarchical or flat.
In a hierarchical information structure, information is organized by higher-order categories, whereby units of information within each category have strong conceptual relationships, but those between categories have weak conceptual relationships. In a flat information structure, information is presented without higher-order categories and units of information have weak conceptual relationships with each other.
To illustrate, an information set that includes ‘cat’, ‘dog’, ‘cow’, ‘mouse’ and ‘tiger’ is hierarchically organized under the higher-order category of ‘animal’. On the other hand, a set of information such as ‘pudding’, ‘Ukraine’, ‘cheque’, ‘mouse’ and ‘symphony’ has a flat information structure, because these terms do not share any obvious higher-order category. We posited that the latter would lead to higher levels of creativity, mainly due to the ‘cognitive flexibility’ it engenders. Cognitive flexibility is the extent to which individuals can easily switch their focus between different categories or perspectives — making it more likely that they will integrate distal information in unique ways. Previous studies have found a positive relationship between cognitive flexibility and creativity.
We felt that in a hierarchical information structure, the presence of a higher-order category influences the interpretation of the information in that category, reducing the possibility for alternative uses of the information; and that in a flat information structure, the absence of higher-order category allows individuals to discover alternative interpretations of the information and increases cognitive flexibility.
In addition, we felt that a flat information structure might actually increase cognitive flexibility, because it introduces higher probabilities of making distal connections between concepts. By definition, the flat information structure has a flat associative hierarchy, meaning that each unit of information has approximately equal probabilities of being next to any other units of information in the set. Therefore, compared to those in the hierarchical information structure condition, individuals presented with a flat information structure may be more likely to discover serendipitous associations between distal information.
Given that our conscious imagination is bounded and our ability to associate distal categories (i.e., cognitive flexibility) is limited, serendipity can refresh habitual thinking and open up new possible associations. History provides numerous instances where serendipitous discoveries — such as the Archimedes principle or the X-ray — have enriched our lives. Based on these previous findings, we felt that a flat information structure would increase the chances for serendipitous, flexible uses of information.
We tested our predictions in three experiments, using a sentence construction task and a LEGO task. In the sentenceconstruction task, participants were given a set of words and asked to construct meaningful sentences out of the them. In the LEGO task, subjects were asked to construct an alien figure from a set of LEGO bricks. Both tasks involved assembling components, but there was no one ‘correct’ way of doing things, and hence efficiency was not a relevant criterion.
One-hundred-and-sixty undergraduate students voluntarily participated in this experiment in exchange for one course credit. Upon arrival, participants were randomly assigned to either a hierarchical information condition or a flat information structure condition. Each received two sheets of letter-size paper and a pencil; the first sheet contained 100 nouns, and the other sheet was for writing down sentences.
In the hierarchical condition, participants were provided with a sheet of letter-size paper containing the 100 English nouns organized by 20 categories. Each category contained five nouns that were conceptually related. We did not provide specific names for each category. In the flat information structure condition, the same 100 English
We felt that a flat information structure might actually increase cognitive flflexibility.
nouns were presented on a sheet of paper without any categorization. Participants were instructed to generate as many sentences as they could by combining the nouns, taking as much time as they needed within a 60-minute time limit.
To evaluate levels of creativity in the sentences generated, three undergraduates from the Linguistics department of a North American university were recruited. They evaluated the level of creativity of each sentence from 1 (not at all creative) to 7 (extremely creative).
In Study 1, participants presented with disorganized RESULT: information — a flat information structure — were more creative than those presented with information organized by categories. The beneficial effect of the flat information structure on creativity was mediated by cognitive flexibility.
For Study 2, in exchange for a course credit, 117 undergraduate students were recruited. We used the same task as in Study 1; the only difference was that we provided 45 English nouns instead of 100. Participants were randomly assigned to the conditions and received a sheet of paper containing 45 nouns either organized by nine categories or unorganized. They were asked to construct as many sentences as they could by combining provided nouns. Unlike Study 1, which used a paper-and-pencil survey, these participants entered their sentences online.
To evaluate the sentences created, we recruited three PHD candidates to evaluate (1) creativity (‘how creative is this sentence?’); and (2) creative use of the provided nouns (‘how creatively has the participant used the provided noun(s) in this sentence?’).
Study 2 supported our hypotheses. Participants in RESULT: the flat-information structure condition generated more creative sentences than those in the hierarchical information structure condition using different measures of creativity. In addition, the effects of information structure on creativity were mediated via cognitive flexibility such that those in the former condition used nouns from more cat- egories than those in the latter condition.
In Study 3, we attempted to show that our predictions are not limited to abstract constructs, but apply to an instance that involves combining units to create new objects. LEGO bricks are analogous to units of information in many ways and, similar to information, there are almost an infinite number of alternative combinations of LEGO bricks.
Just like new information can be created by combining existing information, LEGO bricks can be combined to make complex shapes and structures (e.g., houses, robots, and creatures). Also, similar to other declarative information, LEGO bricks can be categorized by higher-order categories such as colour and shape.
In exchange for one course credit or ten dollars, 182 undergraduate students voluntarily participated in this experiment. We used the Alien Task developed by University of Alabama Professor Thomas B. Ward. Originally, this task asked participants to imagine that they are visiting a planet in another galaxy and encounter an alien who lives on that planet—and to draw that alien. Instead of drawing an alien, we asked participants to build one out of LEGO bricks.
Upon arrival, participants were randomly assigned to either the hierarchical or flat information structure condition. In the hierarchical condition, a total of 442 LEGO bricks, which consisted of nine colours and 11 quadrangular shapes (99 possible categories), were categorized into two large boxes. Each box had 24 cells (four rows, six columns) partitioned by plastic walls. The bricks were categorized by a total of 48 cells. In the flat information structure condition, the same LEGO bricks were contained in two large boxes of identical size to those in the hierarchical information structure condition. However, in the flat condition there was no partition, so all the 442 bricks were mixed and divided into the two large boxes.
Participants were instructed not to pour the bricks onto the table and could only take pieces directly from the boxes when they needed them. Three independent raters were recruited to evaluate the creativity of the LEGO aliens.
Participants in the flat-information condition made RESULT: more creative alien figures than those in the hierarchical condition. Further, the effects of structure on creativity were mediated by both cognitive flexibility and persistence.
Implications for Managers
Across three studies, we found that individuals presented with a flat information structure were more creative compared to those presented with a hierarchical information structure. We also confirmed that the increased creativity in the flat information structure conditions was due to the resulting increased level of cognitive flexibility.
Organizational studies have had a long history advocating for hierarchical structures that increase the efficiency of work. This makes sense, because the processing of large amounts of information is limited by human capacity. As a result, structures have emerged to reduce complexity and enhance efficiency. For this reason, many organizational activities are built around hierarchical structures.
The principle of division of labour, for example, organizes labour forces by worker specializations, creating a hierarchical information structure where information clusters around job roles and skills. Such structures are important and are often necessary to promote efficiency.
However, when it comes to creativity, rigid walls between categories in hierarchical information structures may be harmful, because creative ideas often rise from combining of distal information. Put simply, a hierarchical information structure seems to prime the ideation process within particular cognitive categories, whereas a flat information structure frees up flexible exploration over distal cognitive categories.
Researchers have suggested that functional diversity in teams increases creativity, because team members from various functional and cultural backgrounds bring different repertoires of information to team ideation processes. Our findings provide clues for better managing functionallydiverse teams.
To reap the benefits of a cross-functional team, our research suggests that managers might need to create a flat information structure. This can be achieved by formal procedures such as having team members jot down as many new ideas as possible and then mixing them up (so as not to make any organizing categories salient) before combining them for new product development.
As indicated herein, in addition to making information available to employees, how that information is presented is a critical factor that affects creative output.
di sorganiz participants participants presented with disorganized information were more creative.