WELCOME TO OUR PRESENTATION
If the video does not load automatically, you can find it here: Collective intelligence and idea competitions
We have decided to see the current situation as (a) an opportunity to try a new presentation format and (b) a case study for our presentation. Hence, this storyboard illustrates our algorithm in a way that is …
structured. Thanks to R Markdown, the text of each slide is embedded with the code the algorithm. So you are seeing how the data gets created.
dynamic. We use a clustering algorithm that selects ideas randomly. In the original version, the results change every time you restart the page and that shows that we did not use a specific set of parameters to obtain good results. In this static version, the results do not change over time.
At the end of the dashboard, we share our dictionary and the dataset of quotes, to help you understand how we assessed the ideas and to give you something to read in this special time. We hope you enjoy watching it as much as we enjoyed making it.
01) INTRODUCTION: How to filter ideas while taking into account the seeker’s goals and the learning dynamics?
Ideas competition is an online idea competition initiated by a firm that sets a medium amount of specificity concerning the required tasks, which can range from ideas, sketches, concepts, prototype, and working solutions, for a specific target group of individuals that competes and cooperates thanks to community functions in the online platform, for a large amount of time, in exchange of a mix between monetary and non-monetary rewards that are given by a jury of experts (source).
In recent years, the amount of Google searches for idea challenges such as Innocentive has been intervined with the amount of searches for collective intelligence. We present the preliminary results of a scalable processes that uses idea clustering to (1) reduce the cost per idea selection, (2) increase the quality of the retained ideas, since it is not biased by the bounded rationality of the solution seeker.
02) GAP IN THE LITERATURE REVIEW: Focus on learning as an incentive and as a processs
Most of previous researches has considered the idea provider as main subject of analysis. Instead, we focus on the bounded rationality of the solution seeker and the learning dynamics.
Seekers often estimate that the main cost of an idea challenge comes from the reward itself; yet, making mistakes in the selection process and picking the wrong idea might result in the seeker wasting time and money.
The learning dynamics at the seeker side, can be summarized through the metaphor of a person looking for the right configuration of battery power (consumption of resources) and signal bandwidth for its mobile (different degrees of learning due to the diversity of sources accessed for the ideas).
WIKISURVEY Part 1: The data
| Which.do.you.think.is.better.for.creating.a.greener..greater.New.York.City. | Score..0…100.. |
|---|---|
| Continue enhancing bike lane network, to finally connect separated bike lane systems to each other across all five boroughs. | 63 |
| Create a network of protected bike paths throughout the entire city | 63 |
| Promote cycling by installing safe bike lanes | 62 |
| Require all big buildings to make certain energy efficiency upgrades | 61 |
| Promote the use of solar energy using the latest technology on all high-rise buildings. | 61 |
| Utilize NYC Rooftops to install Solar PV panels | 61 |
| Which.action.do.you.think.the.community.should.prioritize.to.reduce.energy.use.and.greenhouse.gas.emissions. | Score..0…100.. |
|---|---|
| Support micro grid neighbourhood energy sharing (e.g., heating, solar, wind) | 70 |
| Improve public transit infrastructure | 68 |
| Develop solar/wind farms | 67 |
| Support local food availability | 67 |
| Develop neighbourhood energy plans (addressing how energy can be saved and/or generated at the neighbourhood level) | 62 |
| Expand the landfill gas collection system to reduce the gas impact on climate change by about 30X | 61 |
Wikisurvey allows participants to share ideas and collect feedbacks.
The data of this idea competition can be found at this link.
As we can see, the maximum score is below 70. Hence, we normalize the data to obtain scores between 0 and 1.
WIKISURVEY Part 5: Can we predict the score of an idea ?
Comments
STEP 01: Adding ideas from previous competitions
WIKISURVEY Part 3: Dominant topics in winning and loosing ideas
We have asked the system to create … topics.
As we can see, ideas in each topic have received good and bad scores. Hence, the difference of scores between topics in not statistically significant. Nonetheless, there are some topics that are associated to higher scores.
Ideas with high scores talk about Bike (topics 20 and 27) and community (topics 1, 12, 29).
Ideas with high scores talk about price (topics 5 and 10) and street/sidewalk (topics 13, 17).
EXTRA: TIDYMODELS
WIKISURVEY Part 6: Can the model made for competition 01 predicts the score of ideas in competition 02 ?
By following the KDD method, we execute five steps:
- Step 01: We collect the data from competition 02. A sample of the data is shown here.
Step 02: We pre-process the data by converting the ideas into a corpus and the a document frequency matrix
Step 03: We manipulate the data, by using the ideas from competition 01 as training set and the ideas from competition 02 as testing set.
Step 04: We use a random forest algorithm to predict the score of the ideas in competition 02.
Step 05: We illustrate the results, which shows a precision of about 40%.
STEP 00: Injecting new ideas of Competition 01 in Competition 02
Here we assess if competitions on similar topics might have same scoring criteria.
If similar competitions have similar criteria, that would support the idea of a scoring software as a decision support system.
WIKISURVEY Part 7: How would the idea of the first competition score in the second competition ? (cheatstorming)
The problem come from the predictions of C2: 0.5 or 0.75Here we test the notion of cheatstorming: taking ideas from one competition and using them in another one.
Since the competition already took place, we cannot know for sure the score of the new ideas.
This is what we are going to do to guess it: - we build a model to predict the scores of the competition 02 - we test if the model is realiable for competition 02 - we give a score for ideas of competition 01, by using the model for competition 02