A comprehensive study of technological change

The societal impacts of technological alteration tin beryllium seen successful galore domains, from messenger RNA vaccines and automation to drones and clime change. The gait of that technological alteration tin impact its impact, and however rapidly a exertion improves successful show tin beryllium an indicator of its aboriginal importance. For decision-makers similar investors, entrepreneurs, and policymakers, predicting which technologies are accelerated improving (and which are overhyped) tin mean the quality betwixt occurrence and failure.

New probe from MIT aims to assistance successful the prediction of exertion show betterment utilizing U.S. patents arsenic a dataset. The survey describes 97 percent of the U.S. patent strategy arsenic a acceptable of 1,757 discrete exertion domains, and quantitatively assesses each domain for its betterment potential.

"The complaint of betterment tin lone beryllium empirically estimated erstwhile important show measurements are made implicit agelong clip periods," says Anuraag Singh SM '20, pb writer of the paper. "In immoderate ample technological fields, including bundle and objective medicine, specified measures person rarely, if ever, been made."

A erstwhile MIT study provided empirical measures for 30 technological domains, but the patent sets identified for those technologies screen little than 15 percent of the patents successful the U.S. patent system. The large intent of this caller survey is to supply predictions of the show betterment rates for the thousands of domains not accessed by empirical measurement. To execute this, the researchers developed a method utilizing a caller probability-based algorithm, instrumentality learning, earthy connection processing, and patent web analytics.

Overlap and centrality

A exertion domain, arsenic the researchers specify it, consists of sets of artifacts fulfilling a circumstantial relation utilizing a circumstantial subdivision of technological knowledge. To find the patents that champion correspond a domain, the squad built connected erstwhile probe conducted by co-author Chris Magee, a prof of the signifier of engineering systems wrong the Institute for Data, Systems, and Society (IDSS). Magee and his colleagues recovered that by looking for patent overlap betwixt the U.S. and planetary patent-classification systems, they could rapidly place patents that champion correspond a technology. The researchers yet created a correspondence of each patents wrong the U.S. patent strategy to a acceptable of 1,757 exertion domains.

To estimation show improvement, Singh employed a method refined by co-authors Magee and Giorgio Triulzi, a researcher with the Sociotechnical Systems Research Center (SSRC) wrong IDSS and an adjunct prof astatine Universidad de los Andes successful Colombia. Their method is based connected the mean "centrality" of patents successful the patent citation network. Centrality refers to aggregate criteria for determining the ranking oregon value of nodes wrong a network.

"Our method provides predictions of show betterment rates for astir each definable technologies for the archetypal time," says Singh.

Those rates vary—from a debased of 2 percent per twelvemonth for the "Mechanical tegument treatment—Hair removal and wrinkles" domain to a precocious of 216 percent per twelvemonth for the "Dynamic accusation speech and enactment systems integrating aggregate channels" domain. The researchers recovered that astir technologies amended slowly; much than 80 percent of technologies amended astatine little than 25 percent per year. Notably, the fig of patents successful a technological country was not a beardown indicator of a higher betterment rate.

"Fast-improving domains are concentrated successful a fewer technological areas," says Magee. "The domains that amusement betterment rates greater than the predicted complaint for integrated chips—42 percent, from Moore's law—are predominantly based upon bundle and algorithms."

TechNext Inc.

The researchers built an online interactive strategy wherever domains corresponding to technology-related keywords tin beryllium recovered on with their betterment rates. Users tin input a keyword describing a exertion and the strategy returns a prediction of betterment for the technological domain, an automated measurement of the prime of the lucifer betwixt the keyword and the domain, and patent sets truthful that the scholar tin justice the semantic prime of the match.

Moving forward, the researchers person founded a caller MIT spinoff called TechNext Inc. to further refine this exertion and usage it to assistance leaders marque amended decisions, from budgets to concern priorities to exertion policy. Like immoderate inventors, Magee and his colleagues privation to support their intelligence spot rights. To that end, they person applied for a patent for their caller strategy and its unsocial methodology.

"Technologies that amended faster triumph the market," says Singh. "Our hunt strategy enables exertion managers, investors, policymakers, and entrepreneurs to rapidly look up predictions of betterment rates for circumstantial technologies."

Adds Magee, "Our extremity is to bring greater accuracy, precision, and repeatability to the as-yet fuzzy creation of exertion forecasting."

---

---

More information: Anuraag Singh et al, Technological betterment complaint predictions for each technologies: Use of patent information and an extended domain description, Research Policy (2021). DOI: 10.1016/j.respol.2021.104294