Determining general causation from processing scientific articles
Examples of the disclosure are directed toward generating a causation score with respect to an agent and an outcome, and projecting a future causation score distribution. For example, a causation score may be determined with respect to a hypothesis that a given agent causes a given outcome, and the score may indicate the acceptance of that hypothesis in the scientific community, as described by scientific literature. A future causation score distribution, then, may indicate a probability distribution over possible future causation scores, thereby predicting the scientific acceptance of the hypothesis at some specific date in the future. A future causation score distribution can be projected by first generating one or more future publication datasets, and then determining causation scores for each of the one or more future publication datasets.
1. A computer-implemented method of displaying, on a display device, a general causation visualization for an agent and an outcome, the method comprising:
processing a corpus of scientific article metadata to obtain two subsets associated with the agent and the outcome, including a first dataset associated with a first year, and a second dataset associated with a second year;
displaying the general causation visualization on the display device;
displaying, in a portion of the general causation visualization associated with the first year, a representation of a first causation score computed by:
determining a respective magnetism score for each respective article in the first dataset based on directionality data, the directionality data indicating whether the respective article supports or rejects a hypothesis that the agent causes the outcome, and evidence data, the evidence data indicating how well methodology of the article can demonstrate a causal relationship between the agent and the outcome,
aggregating the respective magnetism scores for the articles in the first dataset to obtain a magnetism score for the first dataset,
determining a proximity score by aggregating respective proximity categorizations of each article in the first dataset, each respective proximity categorization indicating directness of evidence in each respective article,
weighting the magnetism score based on the proximity score, and
computing the first causation score based on the weighted magnetism score; and
displaying, in a portion of the general causation visualization associated with the second year, a representation of a second causation score computed based on the second dataset associated with the second year.
2. The method of claim 1 , wherein the first dataset further includes magnitude data, the magnitude data indicating strength of association between the agent and the outcome as observed in an associated article, the method further comprising:
determining a magnitude score based on the magnitude data, wherein the first causation score is further based on the magnitude score.
3. The method of claim 2 , wherein the magnitude data includes an odds ratio, and determining the magnitude score includes multiplying the odds ratio by another value.
4. The method of claim 1 , wherein the first dataset further includes literature impact data, and the magnetism score is further based on the literature impact data.
5. The method of claim 4 , wherein the literature impact data is determined independent of the agent and the outcome.
6. The method of claim 1 , further comprising:
determining a coherence score based on the directionality data and the proximity data, wherein the first causation score is further based on the coherence score.
7. The method of claim 6 , the method further comprising:
calculating a test statistic of aggregated proximity categorizations for articles in the first dataset, wherein the coherence score is determined based on the calculated test statistic.
8. The method of claim 1 , wherein aggregating the respective magnetism scores for the articles in the first dataset includes aggregating the product of respective evidence data and respective directionality data for each respective article in the first dataset.
9. The method of claim 1 , wherein the evidence data includes a categorization of a methodology of the respective article, the method further comprising:
selecting an evidence data value associated with the categorization.
10. The method of claim 1 , wherein each respective proximity categorization categorizes the respective article as at least one of a human study, an animal study, and an in vitro study.
11. The method of claim 1 , wherein the general causation visualization includes a graph of causation score as a function of time, the graph including the representation of the first causation score and the representation of the second causation score.
12. The method of claim 1 , wherein processing the corpus of scientific article metadata includes:
obtaining the first dataset including article metadata within a time threshold of the first year; and
obtaining the second dataset including article metadata within the time threshold of the second year.
13. A non-transitory computer readable storage medium storing instructions executable to perform a method of displaying, on a display device, a general causation visualization for an agent and an outcome, the method comprising:
processing a corpus of scientific article metadata to obtain two subsets associated with the agent and the outcome, including a first dataset associated with a first year, and a second dataset associated with a second year;
displaying the general causation visualization on the display device;
displaying, in a portion of the general causation visualization associated with the first year, a representation of a first causation score computed by:
determining a respective magnetism score for each respective article in the first dataset based on directionality data, the directionality data indicating whether the respective article supports or rejects a hypothesis that the agent causes the outcome, and evidence data, the evidence data indicating how well methodology of the article can demonstrate a causal relationship between the agent and the outcome,
aggregating the respective magnetism scores for the articles in the first dataset to obtain a magnetism score for the first dataset,
determining a proximity score by aggregating respective proximity categorizations of each article in the first dataset, each respective proximity categorization indicating directness of evidence in each respective article,
weighting the magnetism score based on the proximity score, and
computing the first causation score based on the weighted magnetism score; and
displaying, in a portion of the general causation visualization associated with the second year, a representation of a second causation score computed based on the second dataset associated with the second year.
14. The non-transitory computer readable storage medium of claim 13 , wherein the first dataset further includes magnitude data, the magnitude data indicating strength of association between the agent and the outcome as observed in an associated article, the method further comprising:
determining a magnitude score based on the magnitude data, wherein the first causation score is further based on the magnitude score.
15. The non-transitory computer readable storage medium of claim 14 , wherein the magnitude data includes an odds ratio, and determining the magnitude score includes multiplying the odds ratio by another value.
16. The non-transitory computer readable storage medium of claim 13 , wherein the first dataset further includes literature impact data, and the magnetism score is further based on the literature impact data.
17. The non-transitory computer readable storage medium of claim 16 , wherein the literature impact data is determined independent of the agent and the outcome.
18. The non-transitory computer readable storage medium of claim 13 , the method further comprising:
determining a coherence score based on the directionality data and the proximity data, wherein the first causation score is further based on the coherence score.
19. The non-transitory computer readable storage medium of claim 18 , the method further comprising:
calculating a test statistic of aggregated proximity categorizations for articles in the first dataset, wherein the coherence score is determined based on the calculated test statistic.
20. The non-transitory computer readable storage medium of claim 13 , wherein aggregating the respective magnetism scores for the articles in the first dataset includes aggregating the product of respective evidence data and respective directionality data for each respective article in the first dataset.
21. The non-transitory computer readable storage medium of claim 13 , wherein the evidence data includes a categorization of a methodology of the respective article, the method further comprising:
selecting an evidence data value associated with the categorization.
22. The non-transitory computer readable storage medium of claim 13 , wherein each respective proximity categorization categorizes the respective article as at least one of a human study, an animal study, and an in vitro study.
23. The non-transitory computer readable storage medium of claim 13 , wherein the general causation visualization includes a graph of causation score as a function of time, the graph including the representation of the first causation score and the representation of the second causation score.
24. The non-transitory computer readable storage medium of claim 13 , wherein processing the corpus of scientific article metadata includes:
obtaining the first dataset including article metadata within a time threshold of the first year; and
obtaining the second dataset including article metadata within the time threshold of the second year.