www.sciencemag.org SCIENCE VOL 330 15 OCTOBER 2010 317 LETTERS edited by Jennifer Sills 332 Diversity from refuse Targeting ecosystem services Speaking many tongues COMMENTARY 322 LETTERS I BOOKS I POLICY FORUM I EDUCATION FORUM I PERSPECTIVES 323 CREDIT: DAVE PLUNKERT/WWW.SPURDESIGN.COM/DP Machine Science: Truly Machine-Aided Science THE PERSPECTIVE BY J. EVANS AND A . Rzhetsky (“Machine science,” 23 July, p. 399) implies that the next Einstein could be a com- puter. During the past centuries, the episte- mological debates on the scientific trends and revolutionary aspects in the scientific break- throughs of every century showed the central- ity of the investigators and their impressive capabilities (and sometimes their luck). The use of computer-aided simulations and analy- sis has been an important part of modern science; technological developments have led to shorter computational times and the oppor- tunity for scientists to develop an increased number of simulations, mathematical models, and scientific calculi. However, despite the effectiveness of computer tools, the centrality of the research- ers has not changed in science. Scientists define the computation of models, the analy- sis of data, and the validation of scientific hypotheses, as well as the guidelines for ad hoc software and large-scale computations. Technology is a useful tool for the scientists, but it cannot solve open problems such as the Riemann Conjecture. “Machine science” could more accurately be consid- ered “machine-aided science.” Science will not fundamen- tally change until a machine can produce results or solve open problems without human direction. The Perspective describes our current scientific methodology, not the beginning of a new sci- entific era. FRANCESCO GIANFELICI Department of Health Science and Technology, Faculty of Engineering, Science, and Medicine, Aalborg University, Fredrik Bajers Vej 7E-4, DK-9220 Aalborg E, Denmark. E-mail: frgianf@hst.aau.dk Machine Science: What’s Missing MUCH IS MISSING IN J. EVANS AND A. Rzhetsky’s Perspective “Machine science” (23 July, p. 399), which needs to be placed within a broader understanding of scien- tific practice. Recent discussions of data- driven science are misrepresented. Far from “conjectur[ing] that hypotheses are obso- lete,” Golub writes that “hypothesis-driven, experimental research should remain cen- tral” (1), and we write that “hypothesis gen- eration and testing are important to science at many points in a wider topography of inquiry” (2). Leroy Hood, a proponent of Machine Science: The Human Side J. EVANS AND A. RZHETSKY (“MACHINE SCIENCE,” PERSPECTIVES, 23 July, p. 399) misrepresent the crucial role played by humans in using computational tools for automated hypothesis generation. There is no doubt, as they argue, that text mining software and networks of concepts such as those captured by bio-ontologies greatly increase researchers’ ability to mine huge masses of data, compare concepts used across scientific fields, and, ultimately, generate new hypotheses (1, 2). However, both the development and the effective use of these tools are strongly dependent on researchers’ understanding of their sci- entific fields and the data being mined. For instance, building bio-ontologies that adequately represent the concepts and relations used within specific communities requires huge efforts of curation (3, 4). In turn, for researchers to use such ontologies for the purposes of discovery, they must understand how the ontolo- gies were constructed and the choices curators have made in selecting the relevant structures and concepts (5). Without such understanding, researchers are likely to misclassify or misinterpret results. It is thus misleading and unhelpful to equate the effective use of computational tools with the ideal of full automa- tion, as Evans and Rzhetsky have done. Rather, researchers using computational tools should be as aware as possible of the human interventions and assumptions built into those systems, so as to be able to interpret the biological signifi- cance of results obtained through these tools and, when needed, to challenge and/or update the assumptions that they incorporate. SABINA LEONELLI ESRC Centre for Genomics in Society, University of Exeter, EX4 4PJ Exeter, UK. E-mail: s.leonelli@exeter.ac.uk References 1. L. D. Stein, Nat. Rev. Genet. 9, 678 (2008). 2. T. Hey, S. Tansley, K. Tolle, Eds., The Fourth Paradigm: Data-Intensive Scientific Discovery (Microsoft Research, Redmond, WA, 2009); http://research.microsoft.com/en-us/ collaboration/fourthparadigm. 3. D. Howe, S. Y. Rhee, Nature 455, 47 (2008). 4. S. Leonelli, Biol. Theory 3, 8 (2008). 5. S. Leonelli, in How Well Do Facts Travel? The Dissemination of Reliable Knowledge, P. Howlett, M. S. Morgan, Eds. (Cambridge Univ. Press, Cambridge, 2010). Letters to the Editor Letters (~300 words) discuss material published in Science in the previous 3 months or issues of general interest. They can be submit- ted through the Web (www.submit2science.org) or by regular mail (1200 New York Ave., NW, Washington, DC 20005, USA). Letters are not acknowledged upon receipt, nor are authors generally consulted before publication. Whether published in full or in part, letters are subject to editing for clarity and space. Published by AAAS on October 14, 2010 www.sciencemag.org Downloaded from